Grids: The Next Generation · 2008. 12. 8. · by Sarah Pearce, University of London, UK 18 The...

European Research Consortium for Informatics and Mathematicswww.ercim.org

Number 59, October 2004

Special :

Grids:The Next

Generation

2 ERCIM News No. 59, October 2004

3 KEYNOTE

by Fabio Colasanti Director-General European Commission Directorate-General for "Information Society"

JOINT ERCIM ACTIONS

4 CoreGRID Kick-off Meetingby Thierry Priol, IINRIA, France

5 8th ERCIM Workshop "User Interfaces for All"by Constantine Stephanidis, ICS-FORTH, Greece

6 DELOS Information Dayby Costantino Thanos, ISTI-CNR, Italy

EUROPEAN SCENE

7 The EU Information Society TechnologiesProgramme Advisory Group – ISTAGby Michael Wilson, CCLRC, UK

NEWS FROM W3C

8 Tim Berners-Lee Knighted by Her MajestyQueen Elizabeth II

8 COPRAS European Project Kick-off Meeting8 Celebrating W3C's Tenth Anniversary9 Multimodal Web Workshop and Seminars9 W3C Spanish Office to promote W3C

Standards in Tour of Spain10 W3C Spanish Office to promote W3C

Standards in Tour of Spain10 Four W3C Workshops this Fall 10 Semantic Web Standardization in Europe:

SWAD-Europe by Dan Brickley, Libby Miller and Kate Sharp

R&D AND TECHNOLOGY TRANSFER

63 Grammar Enables Effective MultimediaSearch Queriesby Fedde van der Lijn, CWI and MenzoWindhouwer, University of Amsterdam, TheNetherlands

64 From Textual Use Cases to BehaviourSpecificationsby Vladimir Mencl, Charles University/CRCIM,Czech Republic

65 GHT*: A Peer-to-Peer System for Metric Databy Michal Batko and Pavel Zezula, MasarykUniversity Brno, Czech Republic; ClaudioGennaro, ISTI-CNR, Italy

66 CWI's FACS Cluster - Facilitating theAdvancement of Computational Scienceby Stefan Blom, Sander Bohte and Niels Nes,CWI, The Netherlands

68 SynDEx: System-Level CAD Software for Optimizing Distributed Real-TimeEmbedded Systemsby Yves Sorel, INRIA, France

69 Large-Scale Simulations of Diffusion in CellBiologyby Ivo F. Sbalzarini, Anna Mezzacasa, AriHelenius and Petros Koumoutsakos, ETHZurich/SARIT, Switzerland

70 Real-Time MIMO Testbed for Next GenerationWireless LANs by Andreas Burg and Helmut Bölcskei, ETHZurich, Swizterland

EVENTS

72 16th European Conference on Artificial Intelligenceby Vicent J. Botti, Universidad Politécnica deValencia/SpaRCIM, Spain

72 Cross-Language Evaluation Forum — CLEF2004by Carol Peters, ISTI-CNR, Italy

74 Announcements77 Book Review78 EURO-Legal

79 IN BRIEF

CONTENTS

Introduction

12 Grids - The Next Generation -Introduction to the Special Themeby Keith Jeffery, CCLRC, UK and PéterKacsuk, SZTAKI, Hungary

16 Net Generation Grids: The Work of theEuropean Commission Expert Groupby Keith Jeffery, CCLRC, UK

Grid Infrastructure:

17 GridPP: A UK Computing Grid forParticle Physics by Sarah Pearce, University of London, UK

18 The CERN openlab for DataGridApplications by Andreas Unterkircher, Sverre Jarp andAndreas Hirstius, CERN, Switzerland

20 Fraunhofer Grid Alliance – From Grid-Based e-Science to IndustrialApplications by Franz-Josef Pfreundt, AnetteWeisbecker, Jürgen Falkner, Thilo Ernst,Uwe Der, Andreas Hoheisel, Klaus-PeterEckert and Wolfgang Ziegler, FraunhoferICT Group, Germany

21 The Production-Level HungarianClusterGrid Initiative by Péter Stefán, Office for NationalInformation and InfrastructureDevelopment, Hungary

23 Effective Aggregation of Idle ComputingResources for Cluster Computing by Bruno Richard and Philippe Augerat,Icatis, France

24 The European Learning GridInfrastructure Integrated Projectby Pierluigi Ritrovato and Matteo Gaeta,Università di Salerno, Italy

26 DILIGENT: A Digital Library Infrastructureon Grid Enabled Technology by Donatella Castellii, ISTI-CNR, Italy

Grid Architecture:

27 The Cost of Quality for Cycle-HarvestingGrids by Chris Kenyon, IBM Research, ZurichResearch Lab, Switzerland

29 JGrid: A Jini Technology-Based ServiceGrid by Zoltan Juhasz, University of Veszprem,and Gergely Sipos, SZTAKI, Hungary

30 The GRASP Project: InnovativeMiddleware for Next Generation Grid-Based Application Service Providers by Theo Dimitrakos, CCLRC, UK

32 Akogrimo — The Grid goes Mobile by Stefan Wesner, High PerformanceComputing Center Stuttgart, Germany;Theo Dimitrakos and Keith Jeffrey, CCLRC,UK

Grid Middleware:

33 A Component-Based SoftwareInfrastructure for Grid Computing by Christian Pérez, INRIA, France

34 Data Access and Integration by Malcolm Atkinson, National e-ScienceCentre, UK

36 Building a Multi-Domain Grid by Freek Dijkstra, University of Amsterdamand David Groep, NIKHEF, TheNetherlands

37 Resource Management in an OpticalGrid Testbed by Helmut Grund and Wolfgang Ziegler,Institute for Scientific Computing andAlgorithms SCAI, Fraunhofer ICT Group,Germany

39 Grid Resource Management usingSoftware Agents by Isaac Chao, Ramon Sangüesa andOscar Oardaiz, Polytechnic University ofCatalonia, Barcelona, Spain

40 Multi-Agent Systems for Efficient Qualityof Service Routing in Grids by Eric Ayienga, Okello-Odongo, Universityof Nairobi; Bernard Manderick, and AnnNowe,Vrije Universiteit Brussel

Grid Programming:

42 Enabling High-Level Access to GridComputing Services by José Miguel Alonso, Vicente Hernández,Germán Moltó , Universidad Politécnica deValencia/SpaRCIM, Spain

43 Trusted Grid Workflows: Assigning‘Reputation’ to Grid Service Providers by Omer F. Rana, Cardiff University, Wales,UK; Julian Padget, University of Bath, LucMoreau, University of Southampton, UK

44 A High-Level Grid ApplicationEnvironment to Grid-Enable LegacyCode by Péter Kacsuk, Zoltán Farkas, TamásBoczkó , SZTAKI, Hungary; Tamás Kiss,Ariel Goyeneche, Thierry Delaitre,University of Westminster, UK

46 JOpera: Visual Composition of GridServices by Cesare Pautasso ETH Zurich,Switzerland

Grid Applications:

47 Grid-enabled Weka: A Toolkit forMachine Learning on the Grid by Rinat Khoussainov, Xin Zuo andNicholas Kushmerick, University CollegeDublin, Ireland

48 The eMinerals Minigrid: AnInfrastructure to Support MolecularSimulation Scientists by Martin Dove, University of Cambridge, UK

50 Virtual Vascular Surgery on the Grid by Peter Sloot and Alfons Hoekstra,University of Amsterdam

51 Photorealistic Visualisation with Grid-Based Technology by Mika Pennanen and Markus Ylikerälä,VTT Information Technology, Finland

53 Solving Optimization Problems withGrid-Enabled Technologies by Enrique Alba and Antonio J. Nebro,University of Málaga/SpaRCIM, Spain

54 Bioinformatics in the Semantic Grid by Kai Kumpf and Martin Hofmann, Institutefor Algorithms and Scientific Computing(SCAI), Fraunhofer ICT Group, Germany

New Projects:

56 CoreGRID: European Research Networkon Foundations, SoftwareInfrastructures and Applications forLarge-Scale, Distributed Grid and Peer-to-Peer Technologies by Thierry Priol, INRIA, France

57 GridCoord – Co-ordinating Europe-wideInitiatives in Grid Research by Marco Vanneschi , Pisa University, Italy

59 TrustCoM - A Trust and ContractManagement Framework enablingSecure Collaborations in DynamicVirtual Organisations by Theo Dimitrakos, Michael WilsonCCKRC, UK, and Santi Ristol , ATOSOrigin, Spain

60 Grid.it – Next Generation Grid Platformsand their Applications by Domenico Laforenza, ISTI-CNR andMarco Vanneschi, Pisa University, Italy

SPECIAL THEME: GRIDS — THE NEXT GENERATION

3

Grids are an emerging technology that promises tochange the lives of individuals, organisations andsociety as profoundly as e-mail and the Web have done

in the past decade. It is perhaps not surprising that the firstGrids were developed for the European Organisation forNuclear Research (CERN), the particle physics laboratory thatgave us the World Wide Web.

By providing everyone with the high-performance computingand knowledge capabilities previously available only to thelargest corporations and laboratories, Grids have an immensepotential to underpin sustainable growth by improving thecompetitiveness of existing industries and by helping usher innew markets and services previously thought impossible. It isexpected that their impact on our quality of life will beprofound, allowing us to monitor and model more successfullyeverything from global climate change to the way cars behavein collisions.

Grids are both a crucial enabling technology for reaching the‘Lisbon strategy’, set in spring 2000, of transforming theEuropean Union into the most competitive, knowledge-basedeconomy in the world by the year 2010, as well a fundamentalbuilding block in the realisation of the European Research Area(ERA).

In light of this significant potential, the European Commissionhas been financing Grid research since early 2000, when thefirst EU-funded projects were launched under the FifthFramework Programme for Research and TechnologicalDevelopment (FP5). The Grid research projects under FP5were focused on both technology development and applicationpilots. In FP6 (2002–2006) a new technology-driven approachto Grid research is being pursued by the EuropeanCommission, substantiated by the longer-term vision of “NextGeneration Grids” (http://www.cordis.lu/ist/grids). Researchchallenges to be addressed for the realisation of this visioninclude the conceptualisation and/or development of 'network-centric Grid operating systems' as new or enhanced fabrics forfuture distributed systems and services, scale-free, adaptive anddependable Grid architectures and middleware required torealise the 'invisible Grid' vision. In addition, research willaddress new or enhanced Grids service middleware empow-ering individuals and organisations to create, provide and use avariety of services, anytime, anywhere, in a transparent andcost-effective way.

A complementary building block to Grid Technology researchare EU-funded activities in the area of 'ResearchInfrastructures' which include the deployment and testing ofGrids over GEANT, the world’s most powerful research

network which linksover 3 000 researchinstitutions acrossEurope.

The budget for GridTechnology research inFP6 is more than doublethat in FP5 – a cleardemonstration of thefield’s strategic impor-tance. The first 12 FP6 IST Grid research projects launched insummer 2004 with total EU funding of 52M will create a'critical mass' of expertise and resources from across Europe.Ambitious in their scope, they aim to deliver, inter alia, newGrid architectures by the end of the decade, to develop genericGrid technologies and to advance significantly our capabiltiesin vital areas such as security, business models, data integra-tion, programming models, collaboration and knowledgediscovery.

One of the largest projects in the portfolio just launched – theNetwork of Excellence CoreGRID – addresses longer termGrid research and will play an essential role in creating thefoundations for next generation Grids. The European ResearchConsortium for Informatics and Mathematics, ERCIM, leadsthis significant endeavour. The action will bring togetherexisting, but fragmented, Grid research communities. By doingso it will strive for durable excellence, integration and co-oper-ation in Grid research across Europe, thus contributing to thecreation of a European Research Area for Grids.

Considering the significant and sustained European investmentin Grid research, Europe’s strengths in this area are well estab-lished. The challenge is now to create more effective routes forindustrial exploitation in order to translate successfullyresearch results into economic benefits. For Europe to capi-talise better on its strengths, it is indispensable that collabora-tion between research organisations, funding bodies andindustry at all levels of the value chain and across nationalborders is reinforced.

More integrated, long-term research visions and effective plansfor their implementation need to be established, taking intoaccount industrial needs and commercial ambitions. This willbe essential for reaping benefits from the promising area ofGrid research, paving the way towards the provision of Grid,software and knowledge services as a utility.

Fabio Colasanti

Grids: A Crucial Technology for Science and Industry

Fabio Colasanti, Director-General

European Commission Directorate-

General for "Information Society"

KEYNOTE


The kick-off meeting of the CoreGRIDNetwork of Excellence (see also articleon page 56), which looks at large-scale,distributed Grid and Peer-to-PeerTechnology, was recently held at CETICin Charleroi (Belgium). Almost all the42 CoreGRID partners were representedby 73 participants. Thanks to the CETICteam, this first meeting ran smoothly.

The first day commenced with awelcome from the Director of CETIC,Pierre Guisset, who along with his teamhad the responsibility of organizing thismeeting. Franco Accordino, projectofficer at the European Commission incharge of CoreGRID, gave the welcomeaddress, describing the EC vision of howto boost excellence in European Gridresearch through sustainable integrationand cooperation, thanks to the Networkof Excellence instrument within FP6. Hehighlighted the importance in the devel-opment of Grid technology ofCoreGRID within the EC strategy.Following this, representatives of eachpartner introduced themselves to thegeneral audience and gave a briefoverview of their partner’s involvementwith the joint program of activity. Themorning session ended with a generalpresentation of CoreGRID by the scien-tific coordinator, Thierry Priol, and a setof presentations of the six scientific areascovered by the network, given by theirrespective scientific leaders: DomenicoTalia (Knowledge and Data Manage-ment), Marco Danelutto (ProgrammingModel), Artur Andrzejak (SystemArchitecture), Ludek Matyska (GridInformation and Monitoring), RaminYahyapour (Scheduling) and VladimirGetov (PSE, Tools and GRID systems).

The afternoon session was devoted to apresentation on management issues byBruno Le Dantec, which covered indetail various aspects of the administra-tive procedures related to the network:budget, reporting, evaluation, fellowship

and mobility programs. This sessionconcluded with a set of meetings of thebodies that govern the network: theGeneral Assembly meeting, chaired byDomenico Laforenza, the ExecutiveCommittee meeting, chaired by ThierryPriol, and the Integration MonitoringCommittee meeting. For the lattermeeting, a call was issued for candidatesto chair this committee, and a vote wastaken soon after the conclusion of themeeting. Ron Perrott, from BelfastUniversity, was elected as the chairmanof the IMC. The first day concluded witha dinner, allowing participants to pursuetheir discussions over Belgian cuisine.

The second day consisted of a set ofparallel sessions, each dedicated to oneof the six scientific areas. The purpose ofthese sessions was to establish a moreprecise set of expected scientific results,a work-plan and task descriptions, and toreview the deliverables and milestones.Each partner involved in these scientificareas gave a short description of its role,and the scientific leader of each areaproposed an agenda for the next six

months. Dissemination actions were alsodiscussed. A summary of the parallelsessions was reported in a plenarysession that afternoon to inform allparticipants of the future activitiesrelated to the six scientific areas. Theafternoon concluded with presentationson the activities related to integration(given by Sergei Gorlatch), spreadingexcellence (Anne Falier) and collabora-tion with other EU-funded grid projects(Wolfgang Ziegler).

Link: http://www.coregrid.net

Please contact:Bruno Le Dantec, ERCIMAdministrative & Financial CoordinatorTel: +33 4 92 38 50 13E-mail: [email protected]

Thierry Priol, INRIA, FranceScientific CoordinatorTel: +33 2 99 84 72 10E-mail: [email protected]

The CoreGRID Network of Excellence, in the area of Grid and Peer-to-Peertechnology had its first meeting at CETIC, Charleroi (Belgium), on 13–14September, 2004.

CoreGRID Kick-off Meetingby Thierry Priol

EUROPEAN SCENE

From left to right: Pierre Guisset (CETIC), Bruno Le Dantec (ERCIM), Franco Accordino

(EC) and Thierry Priol (INRIA/ERCIM).

©C

ET

IC; P

hoto

: Ale

xis

Tam

inia

ux

ERCIM News No. 59, October 2004

JOINT ERCIM ACTIONS

5

In the tradition of its predecessors (heldin Heraklion, Crete, Greece, 30-31October 1995; Prague, Czech Republic,7-8 November 1996; Obernai, France, 3-4 November 1997; Stockholm, Sweden,19-21 October 1998; Dagstuhl,Germany, 28 November-1 December1999; Florence, Italy, 25-26 October2000; and Paris (Chantilly), France, 24-25 October 2002), the 8th ERCIMWorkshop 'User Interfaces for All' aimedto consolidate recent work and to stimu-late further discussion on the state of theart in User Interfaces for All and itsincreasing range of applications in theupcoming Information Society. Thespecial theme of this year’s workshopwas 'User-Centred InteractionParadigms for Universal Access in theInformation Society'. The requirementfor User-Centred Universal Accessstems from the growing impact of thefusion of the emerging technologies andfrom the different dimensions of diver-sity that are intrinsic to the InformationSociety. These dimensions becomeevident when considering the broadrange of user characteristics, thechanging nature of human activities, thevariety of contexts of use, the increasingavailability and diversification of infor-mation, knowledge sources and e-services, the proliferation of technolog-ical platforms, etc. The 8th ERCIMWorkshop 'User Interfaces for All'focused on the new HCI challengesarising from this evolution, and on howthese affect the continuing effortstowards Universal Access in theInformation Society.

Keynote speakers were Prof. BenShneiderman (University of Maryland,Maryland, USA), who presented'Interface Design Strategies to PromoteLearnability for All', and Prof. JennyPreece (University of MarylandBaltimore County, Maryland, USA),

who talked about 'Online Communitiesfor All'.

The workshop attracted the strongestever interest with over 140 submissionsfrom all over the world, covering a widerange of topics that include novel inter-action paradigms and contexts of use,innovative concepts of universal accessi-bility and sociability, new modalities and

dialogue styles, user-centered design inmobile application scenarios, late-breaking empirical results with respectto assessing universally accessible appli-cations, and standardization efforts.Contributions addressed not only tech-nological solutions, but also designparadigms and empirical methods forevaluation, as well as policy develop-ments.

The official Workshop Proceedings(http://www.ui4all.gr/workshop2004/publications/lncs3196-toc.html) will bepublished by Springer as part of LNCS(Lecture Notes in Computer Science)series, in the subline 'State-of-the-ArtSurveys', and will be embedded in the

LNCS digital library. The workshopadjunct proceedings are electronicallyavailable in PDF format athttp://www.ui4all.gr/workshop2004/publications/adjunct-proceedings.html.

From the year 2001 onwards, theERCIM Workshop 'User Interfaces forAll' takes place every two years, in alter-nation with the newly established

'Universal Access in Human-ComputerInteraction' (UAHCI) Conference, alsoheld every two years in the context HCIInternational and affiliated conferences.In 2005, the 3rd UAHCI Conference willbe held in Las Vegas, Nevada, USA, 22-27 July (http://www.hci-international.org/).

Links:8th ERCIM Workshop 'User Interfaces for All':http://www.ui4all.gr/workshop2004/

ERCIM Working Group 'User Interfaces for All': http://www.ui4all.gr

Please contact:Constantine Stephanidis, ICS-FORTH, GreeceTel: +30 2810 391741E-mail: [email protected]

8th ERCIM Workshop "User Interfaces for All"by Constantine Stephanidis

The 8th ERCIM Workshop 'User Interfaces for All' was held in Vienna, Austria, 28-29 June 2004, and Prof. Chris Stary, University of Linz, was the Programme Chairand local organiser. The workshop built on the results of the seven previousworkshops.

Photo taken during the

workshop. Standing: Ben

Shneiderman (first keynote

speaker), Simone Stoiber

(workshop assistant organiser),

Constantine Stephanidis

(Working Group chair) and

Christian Stary (workshop

organiser). Sitting: Noelle

Carbonell (Workshop PC

member) and Jennifer Preece

(second keynote speaker).


JOINT ERCIM ACTIONS

DELOS is a Network of Excellence onDigital Libraries funded by the SixthFramework Programme (FP6) of theEuropean Union. The Network isplanned as a four-year activity and beganin January 2004. It aims to build on theresults of a previous DELOS initiativeunder FP5. The Network is currentlyconstituted by members from 46 researchinstitutions in 13 European countries andis coordinated scientifically by ISTI-CNR and managed by ERCIM.

The Info-Day was opened by CostantinoThanos (ISTI-CNR), the scientific coor-dinator of the Network, who welcomedthe participants (80 attendees) andpresented the main directions ofDELOS, its internal organization intoclusters (work-packages) and the clusterspecific topics of interest. The mainobjective of DELOS is to define andconduct a joint programme of activitiesin order to integrate and coordinate theon-going research activities of the majorEuropean research teams in the field ofdigital libraries for the purpose of devel-oping the next generation digital librarytechnologies. The Network is organizedinto eight Clusters (Work-packages): • Digital Library Architectures• Information Access and Personalization• Audio/Visual and Non-Traditional

Objects• User Interfaces and Visualization• Knowledge Extraction and Semantic

Interoperability• Preservation• Evaluation • Dissemination.

Claude Poliart, the DELOS projectofficer, also welcomed the participantsand outlined the Commission's expecta-tions from the Network. DELOS shouldtackle the fragmentation of Europeanresearch in the digital library area byintegrating a critical mass of expertise

around a joint programme of activitythus spreading excellence and estab-lishing European leadership at the globallevel. The Commission wants to see ajoint management of the knowledgeportfolio between the Network’smembers and a sharing of research plat-forms, tools and facilities. DELOSshould become a world reference for thedigital library community.

Following these two introductory talks,Yannis Ioannidis (University of Athens)presented the results of a recently orga-nized DELOS brainstorming workshop on'Digital Libraries: Future ResearchDirections'. In particular, he presented theDELOS vision for DL as user-centredsystems which function not just as passiverepositories but as active communicationand collaboration tools. He stated that theactual challenge for the DL researchcommunity was to develop powerful,generic digital library managementsystems rather than ad-hoc solutionsmeeting the particular needs of a specificapplication community.

The various workpackage leaders thenpresented their clusters in terms of activ-ities and results so far achieved. So far,in addition to the brainstorming meeting

in Covara, Italy, attended by 24 interna-tionally recognized experts in digital-libraries-related research areas, theNetwork has organized nine scientificevents on a wide range of topics, severalnational and regional events, and aninternational summer school on digitallibrary technologies.

The two important evaluation initiativessupported by the Network, the CrossLanguage Evaluation Forum (CLEF)(see workshop report on page 72) and theInitiative for the Evaluation of XMLRetrieval (INEX), which aim atpromoting research into multilingualinformation retrieval systems and intocontent-oriented XML retrieval, respec-tively, were also presented.

The meeting concluded with a presenta-tion by Vittore Casarosa (ISTI-CNR)illustrating the dissemination and tech-nology transfer activities of the Networkand the research exchange programme.

Link:DELOS Website: http://www.delos.info/

Please contact:Costantino Thanos, ISTI-CNR, ItalyTel: +39 050 3152910E-mail: [email protected]

The DELOS Network of Excellence on Digital Libraries organized an Info-Day inconjunction with the European Conference on Digital Libraries (ECDL'04) in Bath,UK, on 15 September. The purpose was to present DELOS to the wider Europeandigital library research community in terms of its objectives, organization, activitiesand results so far achieved.

DELOS Information Dayby Costantino Thanos

Costantino

Thanos,

DELOS

coordinator,

opens the

Information

Day.

ERCIM News No. 59, October 2004 7

EUROPEAN SCENE

ISTAG has been set up to advise theCommission on the overall strategy to befollowed in carrying out the ISTthematic priority and related activities ofresearch as well as on the orientationswith respect to the European ResearchArea by helping to stimulate the corre-sponding European research communi-ties.

In the context of building an ISTEuropean Research Area, a key role ofISTAG is to reflect and advise on thedefinition and implementation of acoherent research policy in Europe. Thispolicy should ensure the mastering oftechnology and its applications andshould help strengthen industrialcompetitiveness and address the mainsocietal challenges. It is through such anexercise that recommendations on priori-ties and activities of Community-fundedresearch in IST would be drawn.

ISTAG members are appointed as indi-viduals rather than to explicitly representthe interests of a single group, sinceformal representation would result in toolarge a body to reach significant deci-sions. ISTAG has a Chairman and 30members who provide a reasonablecoverage across the EU member andaccession states (including Rumania andBulgaria), companies (eg Nokia),research institutes (eg DFKI GmbH),standards bodies (eg W3C), universities(eg Open University of Catalonia),national funding agencies (eg EPSRC inthe UK) and even politicians (eg amember of the Parliament of Estonia).Among these individuals two aremembers of ERCIM institutes who canpresent the interests of ERCIM: thechairman Prof. Encarnacao of the

Fraunhofer ICT Group in Germany andProf Michael Wilson, CCLRCRutherford Appleton Laboratory, UK.ERCIM also provides all ISTAGmembers with its publications.

The objectives of the IST thematicpriority are not purely to produce thehighest quality research in Europe whichcan be exploited by European compa-nies, but also to meet the broader objec-tives of the European Union establishedby the treaties and Council of Ministers.Any advice which ISTAG offers theCEC is only taken as a recommendationon priorities and can be overruled byother factors. For example, ISTAGcautioned that the implementation of thenew funding instruments in the 6thFramework Programme would notachieve the objective of freeing researchmanagement without more independentfinancial control, but the powers of thecommission under the treaties preventedthem from devolving more financialcontrol to consortia, and the resultinginstruments were implemented as wenow have them. Similarly, ISTAG hasnot strongly supported proposals forEuropean Technology Platforms (ETP)as a suitable instrument to structureresearch and innovation in the ICT areain FP7, although they are supported byindustries with a different structure, suchas aerospace.

ISTAG meets in plenary session quar-terly to establish working groups andreport on their progress. The topics forworking groups are established by thecommission in consultation with thechairman, although as ISTAG gainsconfidence through working togetherthey are beginning to propose topics

themselves (eg to address legal issuespotentially arising from ICT research).

The three topics addressed by workinggroups in 2003 reflected three stages inthe innovation cycle: IST ResearchContent which mainly developed theAmbient Intelligence vision to guide thestructure of future work programmes;the implementation issues of Humanresources, research infrastructures,funding mechanisms and partnership;the exploitation of research results.

In 2004 the topics considered take adifferent approach, considering thetopics for research in terms of a comple-mentary vision of the Grid, but alsoconsidering Grand Challenges to struc-ture research objectives rather than usingvisions and roadmaps alone; to considera particular instrument for promotinginnovation through Experience andApplication Research Centres (EARC);and to consider actions that can be takenas EU-wide initiatives to promote andadvance European research and tech-nology beyond the existing ISTprogramme instruments and workprogramme structure.

For further information, past workinggroup reports and drafts of those underdevelopment are available on the ISTAGweb site.

Link: ISTAG web page:http://www.cordis.lu/ist/istag.htm

Please contact:Michael Wilson, CCLRCTel: +44 1235 446619E-mail: [email protected]

by Michael Wilson

In planning the Information Society Technologies (IST) programme, the EuropeanCommission Directorate-General for "Information Society" takes advice andguidance from many sources. These include consultations with past and presentproject managers, and active researchers through questionnaires and consultationmeetings before major changes to the programme. One of the standing bodieswhich is available to the European Commission for consultation on the ISTprogramme is the IST Advisory Group (ISTAG).

The EU Information Society TechnologiesProgramme Advisory Group – ISTAG

News from W3C

Celebrating W3C's 10th AnniversaryIn 1994, Tim Berners-Lee created the World Wide WebConsortium (W3C) to "Lead the Web to its Full Potential."This year, W3C celebrates its 10th anniversary, and theConsortium is organizing a one-day symposium (1 December)for Members and invited guests to reflect on the progress of theWeb, W3C's central role in its growth, and risks and opportuni-ties facing the Web during W3C's second decade.

The symposium program will consist of sessions about theWeb to the present and into the future.

Discussion will be led by Web pioneers, prominent technolo-gists, industry leaders and press personalities. The day will endwith reflections and projections by W3C Director and theWeb's inventor, Tim Berners-Lee.

Link: http://www.w3.org/2004/09/W3C10.html

Tim Berners-Lee Knighted by Her Majesty Queen Elizabeth IITim Berners-Lee, the inventor of the World Wide Web andDirector of the World Wide Web Consortium (W3C) wasdubbed a Knight Commander, Order of the British Empire(KBE) by Her Majesty, Queen Elizabeth II during anInvestiture in London on Friday, 16 July 2004. The rank of

Knight Commander is the second most senior rank of the Orderof the British Empire, one of the Orders of Chivalry.

"I am humbled by this great honor," stated Sir Timothy. "TheWeb came about through an ongoing collaboration with myfellow inventors and developers worldwide. Everyone in theInternet community should be recognized by this honor."

He continued, "The Web must remain a universal medium,open to all and not biasing the information it conveys. As thetechnology becomes ever more powerful and available, usingmore kinds of devices, I hope we learn how to use it as amedium for working together, and resolving misunderstand-ings on every scale."

While working in 1980 as a consultant software engineer atCERN, the European Particle Physics Laboratory in Geneva,Switzerland, Sir Timothy wrote his own private program forstoring information using the kind of random associations thebrain makes. The 'Enquire' program, which was neverpublished, formed the conceptual basis for his future develop-ment of the Web. Subsequently he proposed a global hypertextproject at CERN in 1989, and by December 1990, the program'WorldWideWeb' became the first successful demonstration ofWeb clients and servers working over the Internet. All of hiscode was made available free on the Internet at large in thesummer of 1991.

Sir Timothy wrote the book "Weaving The Web"(HarperCollins, 1999) which describes the Web's birth andevolution.

Links: http://www.w3.org/People/Berners-Lee/http://www.w3.org/2004/07/timbl_knighted

Pho

to: L

e F

evre

Com

mun

icat

ions

COPRAS European Project Kick-off MeetingCOPRAS (Co-Operation Platform for Research AndStandards) is a three-year Support Action under the EuropeanCommission's Information Society Technologies (IST)Programme, which objective is to improve the interfacebetween research and standards. The members of the COPRASconsortium are the officially recognized European Standardsorganizations — CEN, CENELEC, ETSI, together with TheOpen Group and the World Wide Web Consortium (W3C).

The COPRAS project held its first kick-off meeting on 14October at the CEN/CENELEC Meeting Centre in Brussels,together with a number of other European IST projects fromcall 1 of the 6th Framework Programme (FP6 IST), some ofwhose work COPRAS will try to make into international stan-dards. People from various working groups inside the standardsorganisations that make up COPRAS were present to helpthose European projects develop a plan (a 'Standards ActionPlan') for the standardization of their work and to couple eachof them with a particular standards body and possibly couple tothem to each other as well.

Visit the COPRAS Web site and subscribe to the COPRASpublic mailing list: [email protected] (Subject:subscribe).

Link: http://www.copras.org/

8

W3C Spanish Office to promoteW3C Standards in Tour of SpainThe W3C Spanish Office organizes its first "W3C StandardsTour" around different Universities in Spain from 3 to 26November 2004. The tour aims to make W3C work known, toestablish contacts with researchers at Universities in Spain andto promote the use of W3C Technologies between studentsfrom technical Universities.

‘W3C Standards Tour’ consists of a route with an environ-mental friendly bus, which will make stops of one or two daysat the following Spanish cities: Gijón (Opening: 3 November),Bilbao (4 - 5 November), Zaragoza (8 November), Barcelona(9 - 10 November), Valencia (11 - 12 November), Sevilla (15 -16 November), Madrid (18 - 19 November), Salamanca (22November), A Coruña (24 November) and Oviedo (Closing: 26November).

The bus has access facilities for people with disabilities, 14multimedia equipments, where demos of W3C Technologieswill be available, projection equipment, videoconference andInternet connectivity via satellite. At the same time, confer-ences about W3C Technologies and meetings with researchersare organized at the host Universities.

Additionally, the “First W3C Prize on Web Standardization”will be held during the tour. This national prize is an initiativeof the W3C Spanish Office to encourage W3C Standards useand adoption within Spanish Universities. The prize will awardthe best prototype, selected among ten finalist prototypes that,in an innovative way, integrates any of the W3C Technologiesin the following fields: Semantic Web, Device Independence,Voice and Multimodal Interaction. The winner will receive anapplied research grant to develop a full project based on his/herawarded prototype.

Links: http://www.w3c.es/gira/info/intro.html.enhttp://www.w3c.es/

Latest W3C Recommendation• 8 September 2004: Speech Synthesis Markup Language

(SSML) Version 1.0; Mark R. Walker, Daniel C. Burnett,Andrew Hunt

An exhaustive list of all W3C Technical Reports: http://www.w3.org/TR/

Multimodal Web Workshop and Seminars On 19-20 July, W3C organized a workshop on MultimodalInteraction, under the auspices of the IST Programme's MWebproject.

As the W3C Multimodal Interaction Activity is entering itsthird year, much progress has been made defining the W3CMMI framework, and the workshop was organised as an oppor-tunity for discussing W3C's current plans and for providingfeedback and suggestions for future work. Another goal of theworkshop was looking for feedback from multimodal usercommunities that are currently less well represented in theW3C Multimodal work (eg automotive industry/telematics,home entertainment industry, healthcare, aerospace and manu-facturing) as well as feedback and input from the Multimodalresearch community.

41 people attended the workshop and 19 papers (publicly avail-able) were presented, both from academia and the industry,contributing for a better understanding of each other's experi-ence and requirements. The workshop concluded with separate"break-out" sessions, where participants gathered to discussproblems of more specific interest to them, such as multimodalinteraction on mobile devices or advanced academic research.Many attendants also expressed a need for a multimodal inter-action authoring language.

The organizers are confident that the results of these discus-sions as well as the quality of the papers and presentationsmade this workshop very successful. All papers and slides areavailable online, but the organizers are still working on editingthe minutes of the discussions, for publication in the next fewweeks.

Multimodal Web Seminars in Madrid and BerlinThese two Multimodal Web seminars' objective is to informEuropean research and industry about the results of W3C’sMultimodal current work. They are organized in Berlin 13October 2004 and Madrid 18 November 2004. They will use acombination of speakers from W3C Members and W3C team.Attendance is open to the public.

Links: MMI Workshop page: http://www.w3.org/2004/02/mmi-workshop-cfp.html

MMI Workshop papers: http://www.w3.org/2004/02/mmi-workshop/papers

Multimodal Web Seminar in Berlin, 13 October 2004:http://www.w3c.de/Events/2004/membermeeting.html

Multimodal Interaction Seminar in Madrid, 18 November 2004:http://www.w3c.es/Eventos/2004/Noviembre/MMI/

9

Four W3C Workshops this FallW3C Workshop on Semantic Web for Life Sciences,27-28 October 2004, Cambridge, USAThe W3C is organizing a workshop to discuss emerging andfuture applications of Semantic Web for Life Sciences (SW-LS). Specifically, how can Semantic Web technologies such asthe Resource Description Framework (RDF), Web OntologyLanguage (OWL) and domain-specific standards such as theLife Sciences Identifier (LSID) help to manage the inherentcomplexity of modern life sciences research, enable diseaseunderstanding, and accelerate the development of new thera-pies for disease?

Link: http://www.w3.org/2004/07/swls-cfp.html

W3C Workshop on Constraints and Capabilities for Web Services,12-13 October 2004, Redwood Shores, CA, USAVarious aspects of a Web service may require description. Thisincludes its constraints (eg, "You must use HTTPAuthentication when accessing this service") and its capabili-ties (eg, "I can support GZIP compression"). Likewise, clientsaccessing Web services have constraints and capabilities oftheir own to consider when accessing services. This Workshopis being held to discuss the establishment of a framework forthe expression of such constraints and capabilities and theirassociation using SOAP, WSDL, HTTP, etc.

Link:http://www.w3.org/2004/06/ws-cc-cfp.html

W3C Mobile Web Initiative Workshop,18-19 November 2004, Barcelona, SpainW3C is thinking of launching a 'Mobile Web Initiative'. Thegoal of the Mobile Web Initiative would be to make Webaccess from a mobile device as simple, easy and convenient asWeb access from a desktop device. Initial ideas for achievingthis goal include developing 'best practices' documents,providing support infrastructures for mobile developers, orga-nizing training programs for Web content providers andcreating validation and conformance testing services for Web-access from mobile devices.

Link: http://www.w3.org/2004/09/mwi-workshop-cfp.html

W3C Workshop on Metadata for Content Adaptation,12-13 October 2004, Dublin, IrelandContent metadata is a way to express the meaning of Webcontent and to map between vocabularies of the XHTMLfamily of markup languages, their attributes and texts, and theunderlying conceptual model that they represent. Contentmetadata promises the ability to go beyond sharing simpletagged content to sharing some of the concepts behind the tagsin the content and better device independent Web access.

Link: http://www.w3.org/2004/06/DI-MCA-WS/cfp.html

Semantic Web Standardization in Europe: SWAD-Europeby Dan Brickley, Libby Miller and Kate Sharp

The SWAD-Europe project nears completion. SWAD-Europe(Semantic Web Advanced Development in Europe) is a fivepartner collaborative EU project, part of the InformationSociety Technologies programme within the 5th Framework.The project has considerably contributed to the evolutionof the Semantic Web.

SWAD-Europe originated within W3C's Semantic WebActivity, extending the approach taken in previous SWADwork to focus on collaborative work undertaken alongsideW3C's RDF (now Semantic Web) Interest Group. The projectwas created to complement and feed into 'standards track' workwithin W3C. SWAD-Europe is a collaboration between W3CEurope, hosted by ERCIM, research institutions at CCLRC andthe University of Bristol (ILRT), and the companies StiloInternational Plc and Hewlett-Packard Ltd.

The Original IdeaSWAD-Europe was designed to support W3C's Semantic Webinitiative in Europe, providing targeted research, demonstra-tions and outreach. The aim was to support the development anddeployment of W3C Semantic Web specifications throughimplementation, research and testing activities. This involvedfinding and maintaining a balance between 'in-house' opensource tool development, community building, outreach andevangelism, combined with research and analysis to support andfield-test Semantic Web standards. In practice, this meant thatmore time was spent working with members of the SemanticWeb Interest Group community than on in-house developmentwork, since this approach helped establish a broader base forSemantic Web development than could be accomplished withina single European project. The Interest Group provided a forumthat connected SWAD-Europe (and other SemWeb-themedprojects, from Europe and elsewhere) to the more formalstandardisation activities of the W3C.

SWAD-Europe's work is best understood in the context of thechallenges facing the RDF project in the 1997-2001 period priorto the creation of the Semantic Web Activity at W3C. In RDF'searly years, a number of difficulties needed addressing. At thattime, RDF technology was perceived by some as complex andoverambitious, with poorly formalised specifications. Softwaretools were relatively immature, with toolkit-specific APIs andquery languages, which meant that application code was oftentightly bound to a specific RDF toolkit. The relationshipbetween RDF and XML in terms of both tools and standardswas unclear. Despite the difficulties faced by RDF in thisperiod, there was nevertheless a lot of enthusiasm from earlyadopters of RDF which emerged in the Interest Group. Thedesign of the project was an attempt to find a way to use thestructure of a European project to support both formal standard-isation work at W3C, and the community fora (Interest Groups)

10

which can provide a home for users of Web technology to sharetheir experiences.

The project's initial workplan included work on reviewing anddeveloping tools to access, store, search Semantic Web data,with the goal of motivating and informing future work on RDFquery languages and APIs. Practical case studies and demoswere proposed, especially in ‘near term’ areas such as thesaurussystems and next-generation blogging tools. A fair number ofworkshops were included, to foster developer discussion on crit-ical 'seed applications', eg image annotation, calendaring, socialnetworking and geo/mapping work. In addition to these efforts,the project tried to emphasise the importance not only of stan-dards-based technology, but on technology integration. Weundertook work on linking XML to RDF, on schema languages,SVG, query, XForms and Web Services. In each case, there wasthe same underlying intent: to enter into dialogue with users ofW3C technology and understand how their practical problemsmight be addressed through combing RDF approaches withother, perhaps better understood, standards-based methods. Thegoal, in effect, was to help demonstrate the basics of SemanticWeb as a fast-maturing technology platform. Reports from eachof these activity areas are available from the project Web site.

Before, During and AfterThroughout SWAD-Europe, we sought to build on pre-existingwork, in particular work which was conducted in public foraand was supported by freely available open source implementa-tions. Previous collaborative efforts (by the project partners andmembers of the broader community) were supported throughSWAD-Europe and subsequently fed into more formal activi-ties within W3C. The project's general approach is to utilizeW3C's Interest Group mechanism to bring collaboratorstogether, test their ideas, and present the results of those collab-orations in a form that can have a good prospect of adoptionwithin the Working Group mechanisms of W3C.

What difference did SWAD-Europe make?As always with such projects, some areas proved morerewarding than others, and the emphasis of the project evolvedin response to successes and opportunities. A significantachievement of the project has been in the area of outreach andcommunity-building. The SKOS (Knowledge OrganisationSystems and the Semantic Web) work has helped to re-engagethe digital library/thesaurus community. The ten workshopsheld as part of the project have attracted diverse participantsfrom multiple countries and specialities, and from the research,open source and business communities. Other successful workin the project has included software development, in particularthe leading open source C implementation of RDF inRedland/Raptor; and well-crafted and appealing demonstratorsin the areas of Semantic Blogging and Semantic Portals,showing that Semantic Web applications can be simple, prac-tical and easy. The pragmatic 'walk before we run' focus of theproject was appreciated both by Semantic Web sceptics and byenthusiasts. Project members have also made substantialcontributions to the editing and chairing of the RDF Core stan-dards, and later, helped to establish 'Semantic Web phase two'groups at W3C: the Data Access and the Semantic Web BestPractices and Deployment Working Groups.

The overarching aim of the project was to provide, through allappropriate means, a body of answers to questions that had previ-ously gone unanswered, and to foster grassroots communitieswithin which such concerns are addressed. Amongst its manythemes, SWAD-Europe provided detailed answers to developerquestions about RDF query and storage (analysis of scalabilityissues; query languages; APIs), and human-oriented classifica-tion (SKOS for thesauri, bookmark sharing, semantic blogging,etc.). The project's final workshop was on the theme of FOAF,Social Networking and the Semantic Web, and illustrated someof the strengths of the project, combining presentations fromacademic, commercial and open source perspectives with activecollaborative work on tools and applications.

One lesson from the project is that it is both important andrewarding to provide an environment where members of thelarger community that surrounds W3C can interact and collab-oratively explore the practical issues around Web technologies.The formal work of the W3C is based on small, highly focussedWorking Groups where individuals commit a lot of time tocreating new Web standards. SWAD-Europe's primary contri-bution was to help create a supportive background environmentfor such work, by allowing a much larger number of geograph-ically-dispersed individuals to participate (through email, IRC,workshops and the Web) in the Semantic Web initiative. Theproject was, in the Semantic Web scene, unique in its emphasisof the practical and Web aspects of 'Semantic Web' for a Webdeveloper audience. The support that SWAD-Europe providedto the RDF and Semantic Web Interest Groups was an impor-tant exploratory step towards a model for wider participation inWeb standardisation work, showing that W3C's successfulWorking Group-led approach can be complimented by abroader participation model which allows individualresearchers and implementors to make real contributions to thedeployment of existing standards and to the creation of newones. The challenge for the future is to work towards a Web inwhich all European research efforts contribute to the communi-ties which underpin the evolution of Web standards.

Groups interested in collaborating with W3C Europe on futureSemantic Web-related projects are invited to contact DanBrickley at [email protected].

• Dan Brickley is on the technical staff of W3C, where heserves as chair of the Semantic Web Interest Group and asSWAD-Europe Project Director.

• Libby Miller is the Semantic Web Group Coordinator atILRT, University of Bristol.

• Kate Sharp is Project Manager for SWAD-Europe at ILRT,University of Bristol.

Links: SWAD-Europe reports index http://www.w3.org/2001/sw/Europe/reports/intro.html

SWAD-Europe project homepage: http://www.w3.org/2001/sw/Europe/

SKOS: http://www.w3.org/2004/02/skos/

11


The issue contains 32 papers that can bedivided into the following main areas:• Infrastructure• Architecture• Middleware• Programming• Applications• New projects

For rigorous research in this field, it isnecessary for Grid infrastructure to bedeveloped. Unfortunately, few func-tional Grid infrastructures have so farbeen created in Europe. It is thereforeencouraging to find four papersreporting on working Grid infrastruc-tures, and three others detailing plans forthe creation of infrastructures. Thelargest production Grid infrastructure inEurope is the LCG Grid developed byCERN and other physics institutes incollaboration with the EDG (EuropeanDataGrid) project. This was successfullycompleted in March 2004.

The first paper describes in detail the UKComputing Grid and its relationship withLCG. LCG is the largest functioning gridin the world, with over 5000 CPUs andalmost 4000 TB of storage at more than70 sites around the world. The secondpaper reports on the CERN openlabproject, a collaboration between CERNand industrial partners to further developthe LCG Grid infrastructure using thelatest industrial technologies. While theLCG’s middleware had initially beentargeted solely for the x86 platform, thecurrent goal is to port the middleware toItanium. Another major focus of theproject is the ‘10 Gigabit challenge’, iehow to achieve data transfer rates of 10

Gb/s both within the opencluster and toother sites. Other important issuesinclude virtualization and automatedinstallation of grid nodes. TheFraunhofer Resource Grid has beenrunning since 2002 in Germany andcombines the computing and storageinfrastructure of the six institutes of theFraunhofer Grid Alliance. This infras-tructure has been successfully applied tosolving engineering problems such ascasting process optimization,microstructure simulation and drugdesign. The Hungarian ClusterGrid is aproduction grid that has been operatingsince mid-2002, and currently involves1100 nodes and provides 500 Gflops ofsupercomputing power for Hungarianacademic users. Its novelty is that thecomponent clusters serve educationalpurposes during the day and areconnected to the ClusterGrid duringnights and weekends. The result is aGrid solution with the extremely inex-pensive annual operation cost of 40,000Euro.

Icatis has developed technology similarto the Hungarian ClusterGrid: duringcluster usage peaks, a number of idleuser machines can be aggregated to thecluster. The system is designed to betransparent to the PC owners, and themachines are only used at times whenthey are likely to be idle (nights, week-ends, vacations etc).

Two special-purpose Grid infrastruc-tures are under preparation and arereported in this issue. The first is theEuropean Learning Grid infrastructure,which integrates service-oriented Grid

The topic of Grids was last addressed in Issue 45 of ERCIM News, in April2001. At that time the field was relatively new. In the last few years, however,significant efforts have been made all over Europe to establish productionGrids, in order to research every aspect and potential application of Gridsystems. The goal of the current thematic issue is to show both selectedresults from this research and applications of Grid systems, and to show thegoals of some newly launched national and EU projects.

Grids: The Next Generationby Keith Jeffery and Péter Kacsuk

ARTICLES IN THIS SECTION

technology with semantic technology forthe creation of future learning scenarios.The second is the DILIGENT project,which will result in a digital libraryinfrastructure based on Grid-enabledtechnology. This will allow members ofdynamic virtual organizations to collab-orate by exploiting shared knowledgeand physical resources.

Grid architecture is another excitingsubject and is addressed by four papersin this issue. One of the most traditionaltypes of Grid architecture is the cycle-harvesting grid. Work at IBM Researchhas focused on the performance issues ofsuch architectures. Their performance-

related work led to a definition of qualityof service known as Hard Statistical QoS(HSQ) and an accompanying systemarchitecture that can support commercialservice contracts. The JGrid project is aninteresting and novel approach tobuilding service-oriented Grid architec-tures based on the Jini technology andJava platform of Sun Microsystems. TheJGrid project extends the look-up serviceof Jini with a generic Grid discoverysubsystem and with a compute broker.Interactive computing tasks are executedby compute services, jobs are executedby batch services, and storage servicesallow users to store and retrieve theirdata from anywhere in the world. A

major obstacle to applying Grid tech-nology in industry is the lack of a Grid-based ASP model. The GRASP projectis an industry-driven European researchproject exploring the use of the Gridservices concept for solving thisproblem. To achieve this, GRASP hasdeveloped an architectural frameworkfor Grid-based Application ServiceProvision and a prototype realization ofthis framework. Since mobile communi-cation technology is widely used ineveryday life, its integration with Gridtechnology will radically advance thepervasiveness of Grid computing acrossEurope. Through the integration of thesetwo technologies, the Akogrimo project

12 Grids - The Next Generation -Introduction to the Special Themeby Keith Jeffery and Péter Kacsuk

16 Net Generation Grids: The Work ofthe European Commission ExpertGroupby Keith Jeffery

Grid Infrastructure:

17 GridPP: A UK Computing Grid forParticle Physics by Sarah Pearce

18 The CERN openlab for DataGridApplications by Andreas Unterkircher, Sverre Jarpand Andreas Hirstius

20 Fraunhofer Grid Alliance – FromGrid-Based e-Science to IndustrialApplications by Franz-Josef Pfreundt, AnetteWeisbecker, Jürgen Falkner, ThiloErnst, Uwe Der, Andreas Hoheisel,Klaus-Peter Eckert and WolfgangZiegler

21 The Production-Level HungarianClusterGrid Initiative by Péter Stefán

23 Effective Aggregation of IdleComputing Resources for ClusterComputing by Bruno Richard and Philippe Augerat

24 The European Learning GridInfrastructure Integrated Projectby Pierluigi Ritrovato and Matteo Gaeta

26 DILIGENT: A Digital LibraryInfrastructure on Grid EnabledTechnology by Donatella Castelli

Grid Architecture:

27 The Cost of Quality for Cycle-Harvesting Grids by Chris Kenyon

29 JGrid: A Jini Technology-BasedService Grid by Zoltán Juhász and Gergely Sipos

30 The GRASP Project: InnovativeMiddleware for Next GenerationGrid-Based Application ServiceProviders by Theo Dimitrakos

32 Akogrimo — The Grid goes Mobile by Stefan Wesner, Theo Dimitrakosand Keith Jeffrey

Grid Middleware:

33 A Component-Based SoftwareInfrastructure for Grid Computing by Christian Pérez

34 Data Access and Integration by Malcolm Atkinson

36 Building a Multi-Domain Grid by Freek Dijkstra and David Groep

37 Resource Management in an OpticalGrid Testbed by Helmut Grund and Wolfgang Ziegler

39 Grid Resource Management usingSoftware Agents by Isaac Chao, Ramon Sangüesa andOscar Oardaiz

40 Multi-Agent Systems for EfficientQuality of Service Routing in Grids by Eric Ayienga , Bernard Manderick,Okello-Odongo and Ann Nowe

Grid Programming:

42 Enabling High-Level Access to GridComputing Services by José Miguel Alonso, VicenteHernández, Germán Moltó

43 Trusted Grid Workflows: Assigning‘Reputation’ to Grid Service Providers by Omer Rana, Luc Moreau and JulianPadget

44 A High-Level Grid ApplicationEnvironment to Grid-Enable LegacyCode by Péter Kacsuk, Tamás Kiss, ArielGoyeneche, Thierry Delaitre, ZoltánFarkas and Tamás Boczkó

46 JOpera: Visual Composition of GridServices by Cesare Pautasso

Grid Applications:

47 Grid-enabled Weka: A Toolkit forMachine Learning on the Grid by Rinat Khoussainov, Xin Zuo andNicholas Kushmerick

48 The eMinerals Minigrid: AnInfrastructure to Support MolecularSimulation Scientists by Martin Dove

50 Virtual Vascular Surgery on the Grid by Peter Sloot and Alfons Hoekstra

51 Photorealistic Visualisation withGrid-Based Technology by Mika Pennanen and MarkusYlikerälä

53 Solving Optimization Problems withGrid-Enabled Technologies by Enrique Alba and Antonio J. Nebro

45 Bioinformatics in the Semantic Grid by Kai Kumpf and Martin Hofmann

New Projects:

56 CoreGRID: European ResearchNetwork on Foundations, SoftwareInfrastructures and Applications forLarge-Scale, Distributed Grid andPeer-to-Peer Technologies by Thierry Priol

57 GridCoord – Co-ordinating Europe-wide Initiatives in Grid Research by Marco Vanneschi

59 TrustCoM - A Trust and ContractManagement Framework enablingSecure Collaborations in DynamicVirtual Organisations by Theo Dimitrakos, Michael Wilsonand Santi Ristol

60 Grid.it – Next Generation GridPlatforms and their Applications by Domenico Laforenza and MarcoVanneschi

14

intends to create a commercial operator-oriented architecture and platform thatsupport the deployment of Grid servicesworldwide.

Grid MiddlewareGrid middleware design is a centralfocus of Grid research. Six papers in thisissue address this extremely importantsubject. The first introduces a newcomponent-based middleware approachdesigned as the next generation of Gridmiddleware. The new concept will allowautomatic deployment of softwarecomponents on large-scale distributedsystems and grids while making the bestuse of available resources. The projectinvolves the design of a runtime system,called PadicoTM, to allow componentsto exchange data independently of theunderlying networking technologies.The appropriate data access and integra-tion support in the middleware is animportant challenge. The UK’s OGSA-DAI system provides a framework andset of components operating across gridsor Web services that deliver these mech-anisms reliably. Using this framework,many projects are building an OGSA-DAI and extending its repertoire. Italready handles a range of relationalsystems, XML databases and collectionsof files.

Unsolved security issues such as authen-tication and authorization remain majorobstacles to the industrial implementa-tion of grids. The research area of Gridaccounting and economy systems hasalso been neglected. The NIKHEF andUvA teams are jointly developing acombined authentication, authorizationand accounting (AAA) service tech-nique. Based on this, site-local autho-rization and creation of virtual organiza-tion will become more dynamic andeasier for both users and administrators.A particular advantage of this approachis that rather than starting everythingfrom scratch, they are building on theVirtual Organization ManagementService (VOMS) developed by INFN inItaly. This is one of the rare cases inwhich one project fertilizes another.Unfortunately not many papers reportsuch result transfer between projects.

Resource management is a classic andcentral issue in Grid middleware design

and it is therefore unsurprising that twopapers deal with this area of research.Within the framework of the GermanVIOLA project, new Grid middlewarecomponents will be developed to exploitthe capabilities of the new underlyingoptical network infrastructure. Thesenew components will include aMetaScheduler (allowing co-allocationof compute resources, network resourceswith necessary QoS or other resourceslike visualization devices) andMetaMPICH communication librariesfor distributed execution of parallelapplications using MPI for inter-processcommunication. The GridCat group atUPC is working on integrating Gridtechnology and software-agent tech-nology in order to create an agent-basedGrid Resource Management system(GRM). Grids and agent communitiesoffer two approaches to open distributedsystems. Intelligent cooperation, coordi-nation and negotiation are issues handledby software agents and are required byGrid users and Grid resource providersin order to efficiently inter-operate.Another project working on the mergingof Grid and agent technologies addressesQoS problems. In their proposed archi-tecture, self-interested agents willinteract using electronic market-basedtechniques with the goal of establishingand maintaining a certain level of QoS inthe Grid network.

High-Level Grid Programming ConceptsThe next field covered by this issue isfrequently neglected. Since there is acertain gap between the end-user needsand the access level provided by Gridmiddleware, in many cases end-usersmust learn low-level command inter-faces if they wish to use the Grid.Moreover they must be fully aware ofthe underlying Grid middlewareconcepts and re-learn the concepts andaccess mechanisms every time a newgeneration of Grid middleware is intro-duced.

High-level Grid programming conceptsand supporting tools and portals aretherefore extremely important for end-users, making access to Grid servicesmore convenient. Four papers of thecurrent issue describe concepts, toolsand portals that will make the lives of

end-users much easier. The SpanishGRID-IT project introduced a newmiddleware design that offers an object-oriented, high-level applicationprogramming interface, which simplifiesthe process of remote task execution inGrid deployment. It is intended to givesupport for parameter sweep applica-tions. The user is provided with a high-level API that exposes a very natural andconvenient point of entry to the Gridservices.

The well-known workflow concept facil-itates Grid programming, and is handledby two papers. Both of these emphasizethe potential commercial/business appli-cations of grids based on their workflowconcept. The first paper extends thecapability of workflow systems in Gridenvironments, generally focusing on thecapacity to enable more reliable, trust-worthy and efficient use of Gridresources. One of the biggest obstaclesto widespread industrial utilization ofGrid technology is the existence of largenumbers of applications written inlegacy code, which are inaccessible asGrid services. The second paper in thisarea introduces a new approach knownas GEMLCA (Grid ExecutionManagement for Legacy CodeArchitecture), to deploy legacy codes asGrid services without modifying theoriginal code. Moreover, such legacycode services can easily be applied bythe end-user within the P-GRADEportal, providing visual workflowconstruction facilities. With simple drag-and-drop style programming, users canbuild complex workflows fromGEMLCA legacy code services. Thesecan be run on complex Grid systemssuch as GT-3 grids, while the low-leveldetails of that grid remain completelyhidden to the user.

Another visual service compositionsystem is Jopera, consisting of a visualcomposition language and a set of inte-grated software development tools forcomposing Grid services. Jopera offersan open, flexible platform for Gridservice composition that significantlyfacilitates the work of end-users inaccessing the Grid.

The final and perhaps most importantclass of papers demonstrates that the




Grid really is able to contribute to therealization of e-science. They also showdifferent application areas in which Gridtechnology can provide significantsupport for the fast and efficient solutionof large scientific problems. The particlephysics application has already beenmentioned in the context of existingproduction Grid infrastructures, butother application areas are also covered.

The first paper in this class describeswork on Grid-enabled Weka, a widelyused toolkit for machine learning anddata mining. It is a large collection ofstate-of-the-art machine-learning algo-rithms written in Java. Weka containstools for classification, regression clus-tering, association rules, visualization,and data preprocessing. In Grid-enabledWeka, execution of these tasks can bedistributed across several computers inan ad hoc Grid. The second describes theeMinerals project, one of the NERC e-science testbed projects. The scientificaim is to use Grid technology to advancethe capabilities of molecular-scale simu-lations for the study of environmentalprocesses. The third application area isextremely important and will open newhorizons for medical treatments.Medical simulations and visualizationstypically require computational power ata level usually unavailable in a hospital.The University of Amsterdam recentlydemonstrated Virtual Vascular Surgery

(a virtual bypass opera-tion), in which large-scale simulation andvisualization capabili-ties were offered asservices on a grid.Visualization can befurther improved andmade photorealisticthrough the use of Gridtechnology. The fourthpaper reports that atVTT, Grid-baseddistributed computingwas utilized in order toproduce photorealisticimages in a faster andcheaper way. A newgrid known as Grix wasdeveloped based onplatform-neutral archi-tecture.

Complex problems that can only besolved in non-polynomial time arebecoming common in many domains ofour lives: economy, industrial environ-ments, bioinformatics, etc. For such awide spectrum of problems, heuristicscome to the rescue, since exact tech-niques are unable to locate any kind ofsolution. The fifth paper addresses theseissues, showing current research onsolving optimization problems withGrid-enabled technology like Globus,Condor, Legion and SGE.Bioinformatics has already beenmentioned but the last paper in thisgroup puts it in a new context. Fullyannotated data Grid services have enor-mous relevance in bioinformatics, espe-cially systems biology, due to the vastnumber of biochemical entities andinteractions that need to be kept track of.In order to tackle the problem, an annota-tion tool for data in Grid services hasbeen developed by the BioinformaticsDepartment and the Department for WebApplications of Fraunhofer Institute.

Such a theme as this is incompletewithout a look at plans for creating thenext generation of grids. Four papers arereporting newly launched projects thataddress a variety of aspects of futureresearch. The CoreGRID Network ofExcellence project is probably thelargest and most influential of these inEurope. It commenced activity on 1st

September 2004 with the aim ofallowing European researchers todevelop next-generation Grid middle-ware through a high-level joint programof activities. Another importantEuropean project is GridCoord, which isa Special Support Action project. Thegoal is the coordination of Europeaninitiatives and research activities in Gridcomputing in order to strengthencooperation between agencies planningfuture activities, to enhance collabora-tion between research and user commu-nities, and to develop visionary nationaland EU programs and roadmaps.

Yet another EU project is TrustCom, anintegrated project that aims to develop aframework for trust, security andcontract management in dynamic virtualorganizations. The framework willenable collaborative business processesto be undertaken securely in self-managed and dynamic value-chains ofbusinesses and governments. A fourthpaper describes a national Grid projectfrom Italy. Grid.it has a strong interdisci-plinary character and aims to define,implement and apply innovative solu-tions for networked computing-enablingplatforms, which are oriented towardsscalable virtual organizations.

Finally, we provide a forward-lookingperspective with a short paper on NextGeneration Grids, which describes thework of the Expert Group convened bythe EC and also provides the URLs to theoriginal reports of the group. Consideringthe importance of this subject we posi-tioned this paper as the first article of thisSpecial Theme section.

Naturally, one thematic issue cannotcover the entire spectrum of current andplanned Grid research in Europe.However, we believe that the 32 papersbriefly introduced above give a goodoverview of the major Grid researchdirections in Europe. From these papers,readers will gain an overall picture ofEuropean Grid research, and can posi-tion their own strategies on dealing withGrid systems in the future.

Please contact: Péter Kacsuk, STZAKI, HungaryE-mail: [email protected]

Keith Jeffery, CCLRC, UKE-mail: [email protected]


Simulation of a particle physics experiment. Read about

the Large Hadron Collider Computing Grid in the articles

on pages 17 and 18.

©C

ER

N G

enev

a



The Next Generation GRIDs first expertgroup (NGG1) included also fromERCIM: Seif Haridi (SICS) LudekMatyska (CRCIM) and Jesus Labarta(SPARCIM). The report titled 'NextGeneration Grids, European GridResearch 2005-2010' was produced inthe summer of 2003 and is availableunder http://www.cordis.lu/ist/grids.The key concept was ‘the invisibleGRID’; the GRIDs environment shouldjust ‘be there’ for use by applications inscience, business, healthcare, environ-ment, learning and culture domains.

The FP6 Call2 proposals in the GRIDsarea were evaluated and 4 major —CoreGRID, AkoGrimo, NextGRID andSIMDAT — and 8 smaller projects weresuccessful; ERCIM is well representedthroughout and CoreGRID (see article inthis issue) is coordinated by ERCIM.These projects started officially insummer 2004, synchronized with theEuropean GRID Technology Days,September 2004.

However, by early 2004 it was felt thatprojects emerging from Call2, whilecovering many of the issues identified inthe NGG1 report, did not cover them all.Also, increasing experience of the use ofmiddleware designed for metacom-puting (connected supercomputers) inmany other aspects of IT — eg massivedata handling — had revealed someinadequacies. In particular, despite thewelcome evolution of GRIDs middle-ware towards service-orientedcomputing — with the adoption from theWorld Wide Web Consortium webservices the concept of grid servicesdefined by OGSA — the architecture of

current middleware did not make it easyto address major requirements of NextGeneration Grids such as self-managing,self-repairing, fault-toleration and scala-bility. Clearly a further tranche ofresearch topics needed to be defined.The EC reconvened an expert group(NGG2) with some of the NGG1members (including the original 3 fromERCIM) and some new ones with a briefto produce a new report covering – withthe increased knowledge of a furtheryear – the R&D requirements asperceived now.

The NGG2 report (published September2004 available under http://www.cordis.lu/ist/grids) is quite radical; it proposes anew architecture stack for Grids (seeFigure). This stack has applicationssupported by middleware (like Gridservices but with many new features)and in turn by foundationware on top ofexisting operating systems in order toenhance existing operating systems to acommon interface suitable forsupporting adequately Grids middle-

ware. More dramatically, the reportproposes the development of a newGrids operating system with the featuresof the foundationware — and possiblyalso the middleware — built-in in amodular, component-based fashion sothat for any device (and indeed for anyparticular role played by a device at anyone time) only the components requiredare loaded. The objective is a commonGrids IT surface from RFID tags throughsensors, embedded and control systemcomputers, mobile phones, PDAs,laptops, desktops, departmental serversand supercomputers with massive datastores. In fact, the vision combinesaspects of traditional GRID computing(metacomputing) with both distributedand P2P (peer-to-peer) architectureswith many additional novel features.The key concept is the self-managingGRIDs environment with open standardinterfaces to allow multiple commercialofferings.

The vision requires the discarding ofmany long-cherished computer science

by Keith Jeffery

The newly-formed European Commission Directorate General 'Infomation Society'(EC DG INFSO) Unit F2 organised a workshop to discuss the future of GRIDs inEurope in January 2003. Keynote speakers included three ERCIM researchers:Thierry Priol (INRIA), Domenico Laforenza (CNR) and the author. The workshopwith over 200 attendees was a success, and it was decided to form an expertgroup — including the 3 keynote speakers — to take the ideas forward providinga report to the EC to assist in formulation of the later part of Framework Programme(FP) 6 and the upcoming FP7.

Next Generation Grids : The Work of the European Commission Expert Group

A New Architecture Stack for Grids.



principles. Global state can no longer bemaintained and multiple local states withan interconnection and synchronizationprotocol is required – the so-called ‘softstate’ or ‘relative state’. ACID (atom-icity, consistency, isolation, durability)properties of database transactions uponwhich almost all business depends todayare no longer tenable; it is impossible tomaintain, eg a lock for update acrossmultiple self-managing nodes and sonew transaction models, and new trans-action compensation and recoverymodels, are required. The ‘classical’notions of security are also no longerrelevant; cooperating systems withdifferent security policies and protocolswill have to negotiate trust arrangementsin order to provide end-to-end security

(identification, authentication and autho-rization). Synchronisation and real-timeoperation is a real issue: for some appli-cations it is essential (eg controlsystems) and so special protocols andguaranteed services will be required. Inan environment with millions ofconnected nodes, intermittent andmobile communications and semanticrich messages (with linguistic andcultural heterogeneity) carried within aformally-defined syntax there are plentyof R&D challenges for informationsystems engineering.

This vision will permit the realization ofapplications eg intelligent clothes, theintelligent home, the intelligently-assisted hospital, the remotely-

controlled process plant, advisor systemsfor various aspects of life and business,environmental monitoring and controlsystems, enhanced lifelong e-learning,intelligent financial management,advanced scientific systems withmodeling and activation of detectors,disaster management and many more –all of which are at best only part-realisedwithout GRIDs. More importantly - andsubject to trust, security and privacy -these systems will all interoperate asrequired, and with a common user inter-face structure, to weave the fabric of theknowledge society.

Please contact:Keith G Jeffery, CCLRC, UKE-mail: [email protected]: +44 1235 44 6103

In 2007, CERN will introduce its LargeHadron Collider (LHC) — the world’slargest particle accelerator. LHC willallow scientists to penetrate further intothe structure of matter and recreate theconditions prevailing in the earlyuniverse, just after the Big Bang. But thefour experiments at the LHC willproduce more data than any previouscoordinated human endeavour — 10Petabytes each year, equivalent to astack of CDs twice the height of MountEverest. Careful analysis of all of thiscomplex data will be required to look forsome of the most elusive constituents ofcurrent physics, such as the Higgsparticle and supersymmetry.

To deal with this data, LHC will usedistributed computing. More than100,000 PCs, spread at one hundred insti-tutions across the world, will allow scien-tists from different countries to access thedata, analyse it and work together ininternational collaborations. Even today,

the LHC Computing Grid is the largestfunctioning Grid in the world, with over5,000 CPUs and almost 4,000 TB ofstorage at more than 70 sites around theworld. With up to 5,000 jobs being run onthe LHC Computing Grid (LCG) simul-taneously, it is becoming a true produc-tion Grid.

A Particle Physics Grid for the UKGridPP is the UK’s contribution toanalysing this data deluge. It is a collabo-ration of around 100 researchers in 19UK University particle physics groups,CCLRC and CERN. The six-year, £33mproject, funded by the UK ParticlePhysics and Astronomy ResearchCouncil (PPARC), began in 2001 andhas been working in three main areas:• developing applications that will allow

particle physicists to submit their datato the Grid for analysis

• writing middleware, which willmanage the distribution of computing

jobs around the grid and deal withissues such as security

• deploying computing infrastructure atsites across the UK, to build a proto-type Grid.

The UK GridPP testbed currentlyprovides over 1,000 CPUs and 1,000 TBof storage to LCG, from 12 sites in theUK. It is linked to other prototype Gridsworldwide, and has been tested byanalysing data from US particle physicsexperiments in which the UK is involved.Several other smaller experiments havealso started to use the prototype Grid, andparticle physicists are using it to run ‘datachallenges’, that simulate the data anal-ysis needed when LHC is up and running.In this way, UK particle physics hasprogressed ‘from Web to Grid’.

A Tiered StructureThis GridPP testbed is being developedon a hierarchical model, reflecting theoverall structure of the wider LCG

by Sarah Pearce

UK particle physicists have built a functioning prototype Grid, to analyse the datadeluge from CERN’s next particle accelerator. Over the next three years, this willbe scaled up and integrated further with other Grids worldwide, to produce theworld’s first persistent, international, production Grid.

GridPP: A UK Computing Grid for Particle Physics



testbed. CERN provides the 'Tier-0'centre, where the LHC data will beproduced. GridPP has contributed £5mto the CERN for this, which has beenused to support staff and buy hardware.The UK’s 'Tier-1' centre at RutherfordAppleton Laboratory focuses on datastorage and access. In addition there arefour smaller, regional, 'Tier-2s' in theUK, with a focus on provision ofcomputing power for generating simu-lated Monte Carlo data and for analysisof data by individual physicists. In addi-tion, the Grid Operations Centre (GOC),based at RAL, monitors the operationalstatus of resources deployed internatio-nally through LCG and in the UKthrough GridPP.

GridPP2 – The Next PhaseThe second phase of the GridPP projectbegan on 1 September 2004. In the leadup to 2007, this will extend the UKparticle physics grid to the equivalent of10,000 PCs. The infrastructure in the UKwill be continually tested, both by currentexperiments and by the LHC data chal-lenges to ensure that the final system isready by 2007. By the end of this secondphase of GridPP, UK physicists will beanalysing real data from the LHC, usingthe UK Grid for particle physics.

International CollaborationAs well as working with other interna-tional particle physics experiments,GridPP is playing a leading role inEuropean Grid projects. During its firstthree years, GridPP personnel were inte-gral to the EU-funded DataGrid project,which brought together scientists fromEarth observation, bio-medicine andparticle physics to create prototype aEuropean-wide Grid. By the time of itsfinal review in March 2004, EUDataGrid had produced around a millionlines of code and had a testbed of 1,000

CPUs, which had run more than 60,000jobs.

GridPP is now involved in the follow-onEGEE project (Enabling Grids for E-science in Europe), which aims tosupport the European Research Area bybringing together Grids from differentcountries and different disciplines.

Working beyond Particle PhysicsWithin the UK, GridPP2 is also collabo-rating with other parts of the UK’s e-science programme, such as the NationalGrid Service. Many of the tools developedby GridPP could be useful for other disci-plines – for example, GridPP is workingwith clinical researchers on the potentialfor using its computer security tools in thehealth service. In addition, GridPPmembers are collaborating with industry,sharing experience of current Grid devel-opment issues and solutions adopted.

Links:GridPP: http://www.gridpp.ac.ukThe LCG project: http://lcg.web.cern.ch/LCG/EGEE: http://public.eu-egee.org/PPARC e-Science:http://www.pparc.ac.uk/Rs/Fs/Es/intro.asp

Please contact:Sarah Pearce, GridPP Dissemination Officer Queen Mary, University of London Tel: +44 20 7882 5049 E-mail: [email protected]

In the 1000 million short lived particles of

matter and antimatter are studied each

year in the LHCb particle physics

experiment. In order to design the

detector and to understand the physics,

many millions of simulated events also

have to be produced.

CERN, the European Organization forNuclear Research, is currentlyconstructing the Large Hadron Collider(LHC) which will be the largest andmost powerful particle accelerator everbuilt. It will commence operation in2007 and will run for up to two decades.Four detectors placed around the 27 kmLHC tunnel on the outskirts of Genevawill produce about 15 Petabytes of dataper year. Storing such huge amounts ofdata in a distributed fashion and makingit accessible to thousands of scientists

around the world requires much morethan a single organization can provide.CERN has therefore launched the LHCComputing Grid (LCG), with themission of integrating tens of thousandsof computers at dozens of participatingcentres worldwide into a globalcomputing resource. At the time ofwriting, the LCG provides about 5200CPUs and 7.5 TB storage at 68 sites. Inthis context CERN established the‘CERN openlab for DataGridApplications’, a new type of partnership

between CERN and industry. Its goal isto provide CERN with the latest indus-trial technology in order that it mightanticipate possible future commoditysolutions and adjust its Grid technologyroadmap accordingly. The focus ofopenlab's research is to find solutionsthat go beyond prototyping, and thusprovide valuable input to the LCGproject. In this article we highlight someof our research projects.

The CERN openlab is a collaboration between CERN and industrial partners todevelop data-intensive grid technology to be used by a worldwide community ofscientists working at the next-generation Large Hadron Collider.

The CERN openlab for DataGrid Applicationsby Andreas Unterkircher, Sverre Jarp and Andreas Hirstius


19

Openlab runs a so-called openclustercomprising just over 100 Itanium-2 dual-processor HP rx2600 nodes under Linux.Several machines are equipped with 10GbE Network Interface Cards deliveredby Intel. Enterasys provided four 10-Gps‘N7’ switches and IBM delivered 28TBof storage with its ‘Storage Tank’ filesystem. Openlab actively participates indata challenges run by CERN to simulateall aspects of data collection and analysisover the grid. Currently one challenge inpreparation is the detection of the limitsof GridFTP when sending huge amountsof data between CERN and several majorGrid sites around the world. In additionwe focus on an overall ‘10 Gigabit chal-lenge’, namely, how to achieve datatransfer rates of 10 Gb/s within the open-cluster as well as to other sites. Thisinvolves technological issues likenetwork-switch tuning, Linux kernel andTCP/IP parameter tuning and so on. Witha view towards cluster-based gridcomputing we are evaluating a 12-wayInfiniBand switch from Voltaire. Thispiece of technology allows data to be

streamed into the opencluster veryquickly with minimal loss of CPU cycles.

One of our key goals has been to fullyparticipate in LCG. However LCG'smiddleware had initially been targetedonly for the x86 platform. We thereforehad to port the middleware to Itanium,and for several months this became amajor task at openlab. The LCG soft-ware stack consists of a special patchedGlobus 2.4 version provided by theVirtual Data Toolkit (VDT), middlewaredeveloped by the European Data Grid(EDG) project, and several tools deliv-ered by the high energy physics commu-nity. We are able to provide a fully func-tional LCG grid node on Itaniumsystems including elements such asworker nodes, computing elements, theuser interface and storage elements andresource brokers. Thanks to this effort, anew level of heterogeneity has beenadded to LCG. Being the first commer-cial member of LCG, HP sites at PuertoRico and Bristol (UK) are contributingItanium nodes to LCG with the technicalhelp of openlab staff.

Having different platforms poses newchallenges to grid computing. Scientificsoftware is usually distributed in form ofoptimized binaries for every platformand sometimes even tightly coupled tospecific versions of the operatingsystem. One of our projects at openlab isto investigate the potential of virtualiza-tion within grid computing by using thevirtual machine monitor Xen, developedby the University of Cambridge. A gridnode executing a task should thus be ableto provide exactly the environmentneeded by the application.

Another area of interest is automatedinstallation of grid nodes. Originatingfrom different sources, the installationand deployment of LCG middleware is anon-trivial task. We use the SmartFrogframework, a technology developed byHP Labs and given open-source statusthis year, to automatically install andmanage LCG nodes. Of particularinterest is the possibility of dynamicallyadding or deleting resources (ie workernodes and storage elements) to and froma grid node.

Earlier this year Oracle joined openlab asa sponsor. The first project within thiscollaboration aims at reducing the down-time of LCG's replica catalogue, whichensures correct mapping of filenamesand file identifications. The catalogueruns on the Oracle database. With thehelp of Oracle's technology, cataloguedowntime (eg for maintenance reasons)has been reduced from hours to minutes.

Openlab's core team consists of threeCERN staff members, four fellows andsix summer students. Plans for the futureof openlab include increasing the numberof nodes and upgrading the high-speedswitching environment (both Ethernetand InfiniBand). The software activitieswill continue to focus on participation inthe LHC data challenges as well assupporting LCG on the Itanium platform.

Link: http://cern.ch/openlab

Please contact: Sverre Jarp, IT Department, CERN, SwitzerlandTel: +41 22 767 4944E-mail: [email protected]

CERN openlab core

team and contributors

in June 04.

Computer-

generated image

of the LHC tunnel.

©C

ER

N G

enev

a




In the year 2001 five FraunhoferInstitutes launched a project named ‘I-Lab’ which was funded by theGerman Ministry for Science andEducation (BMBF).

This project developed an Internet labo-ratory software on the basis of aFraunhofer Computing Grid. The objec-tive of the project was to provide a user-friendly computing grid which givesusers access to the grid resources andprovides appropriate solutions to theircomputing problems. The FraunhoferResource Grid (FhRG) emerged fromthese efforts, and substantial applicationshave been running on the FhRG since2002. The FhRG Web portal can bereached at http://www.fhrg.fhg.de.Today, the FhRG combines thecomputing and storage infrastructure ofthe six participating institutes: ITWM(Kaiserslautern), IAO (Stuttgart), FIRST(Berlin), IGD and SIT (both inDarmstadt) and SCAI (Sankt Augustin).

Within the I-Lab project, Fraunhoferdeveloped a middleware package basedon, but not strictly dependent on theGlobus Toolkit. The result of theseefforts was to a large extent madepublicly available in the beginning of2004 and is distributed through theeXeGrid Open Source Project athttp://www.exegrid.net.

The software packages developed for theFraunhofer Resource Grid and which arecontinually being developed further inthe eXeGrid Project cover the areas ofgrid workflow systems, grid workflowediting, scheduling of jobs, resourcebrokerage, an intelligent job executionsystem, a Web portal with user manage-ment, management of grid resource

descriptions, resource search mecha-nisms including the mapping of a user’stasks to suitable grid resources, andfinally, an advanced security conceptincluding the respective implementation.The basis on which all these subsystemswork together is a set of resource andworkflow description languages referredto as the ‘Grid Application DefinitionLanguage’ (GADL), and a componentmodel designed for easy integration oflegacy applications into complex gridworkflows.

The Fraunhofer Resource Grid can beaccessed in a number of ways,depending on the user’s level of experi-ence. Figure 1 gives a brief overviewover the FhRG/eXeGrid architecture andthe information flow within it.

As shown, the user can access all thegrid’s services through the Web portal,then going through a process of ‘taskmapping’ that returns appropriate gridresources to solve the task defined by theuser. After having chosen the resources,

the user opens the Grid Job Builder – agraphical workflow editing tool – fromwithin the portal. With this tool complexgrid jobs can be created, which are thensubmitted for execution by the Grid JobHandler Web service. This Web servicedynamically deploys grid workflows tothe hardware resources that are most suit-able for the execution of jobs at the time.The Grid Job Handler communicateswith resource brokers and schedulers andtranslates the above grid workflow intoGLOBUS commands so it can be run onthe underlying grid infrastructure.

As shown in Figure 1, each level can inprinciple be bypassed, which may behelpful for advanced users or specialapplication scenarios.

In addition to the efforts within the I-Labproject, other institutes in the FraunhoferSociety, namely Fraunhofer SCAI andFraunhofer FOKUS, have been involvedin Grid-related efforts, with SCAIcovering supercomputing problems suchas the development of meta-message

by Franz-Josef Pfreundt, Anette Weisbecker, Jürgen Falkner, Thilo Ernst,Uwe Der, Andreas Hoheisel, Klaus-Peter Eckert and Wolfgang Ziegler

The ICT group of German-based Fraunhofer Society, Europe’s largest researchorganization for applied research, has announced the foundation of the FraunhoferGrid Alliance, a network of institutes that combines the special Grid-related skillsof different institutes within the Fraunhofer Society and extends the FraunhoferResource Grid.

Fraunhofer Grid Alliance – From Grid-Based e-Science to Industrial Applications

Figure 1:

eXeGrid

architecture

and

information

flow.



passing and code-coupling algo-rithms for grid environments andmeta-scheduling for co-allocationof various resources includingnetwork links or visualizationdevices. For topics such as theauthentication, authorization andaccounting of users, the definitionand supervision of service-levelagreements and service qualities,network monitoring, and themobile access to Grid services,synergies can be expected betweenthe telecommunication experienceof FOKUS and the development ofcorresponding solutions in the Gridcommunity. The same is true forresearch in the areas of program-ming models, component tech-nology and autonomic communi-cation.

Together, the institutes mentionedabove now form the ‘FraunhoferGrid Alliance’ (see Figure 2),which coordinates the skills andstrengths of the partner institutes to offercombined grid technology and services topublic and industrial customers.

The members of the Fraunhofer GridAlliance currently participate inEuropean-level Grid projects such asEGEE (Enabling Grids for e-Science inEurope), SIMDAT (Data Grid forProcess and Product Development UsingNumerical Simulation and Knowledge

Discovery), K-Wf Grid (Knowledge-based Workflow System for GridApplications) and the CoreGRIDEuropean Network of Exellence. In addi-tion, the Fraunhofer Grid Alliance isstrongly involved in the German ‘D-Grid’ initiative for Grid-based e-Science.

The individual institutes also use tech-nology and experience gained withinAlliance activities in smaller, regional-

scale projects, often involvingSMEs. As an example, one of themost demanding FhRG pilot appli-cations is the Grid-based environ-mental risk analysis and manage-ment system ERAMAS, which isbeing developed by FraunhoferFIRST together with two environ-ment consulting companies(http://eramas.first.fhg.de).

FhRG technology has already beensuccessfully applied to solve engi-neering problems from a broadrange of disciplines, for example,casting process optimization,micro-structure simulation anddrug design.

Making the Grid and e-sciencevision a reality – not only inresearch, but also in industrialsectors such as engineering and lifescience – will remain the primaryobjective of the Fraunhofer GridAlliance in years to come.

Links: Fraunhofer Resource Grid Portal: http://www.fhrg.fhg.deFraunhofer ICT Group: http://www.iuk.fraunhofer.de/

Please contact: Jürgen Falkner Institute for Industrial Engineering IAO, Fraunhofer ICT Group, GermanyE-mail: [email protected]

Figure 2: Fraunhofer Grid Alliance – locations

connected in the Fraunhofer Resource Grid.

The Hungarian ClusterGrid Infrastructurehandles the Grid problem in a slightlydifferent way to contemporary gridsystems, building up from low-levelhardware to the application level by

introducing a layered architecturalmodel. As its name suggests, the basicbuilding blocks of the grid are PC labsthat perform dual functions. During theday, the labs, which are located at

universities, high schools or even publiclibraries, serve an educational purpose inan office-like environment. Wheneverthey are fulfilling this function – typi-cally during nights and weekends – they

by Péter Stefán

A breakthrough represented by the Hungarian ClusterGrid was achieved whenproduction grid development commenced in mid-2002, involving hundreds ofdesktop computer nodes. Launching ClusterGrid has been a result of centredresearch/development and organizational/management efforts. However, theinvestments have already returned manifold by using the 500 Gflopssupercomputing power of a production PC grid in many scientific fields.

The Production-Level Hungarian ClusterGrid Initiative



are used for supercomputing in a Unixenvironment.

The production grid is used byresearchers from many scientific fieldsand their industrial partners: mathe-matics, biology, chemistry, and datamining are a few examples of scientificareas that need large amounts of compu-tational power and make use of such

facilities with true parallel and parameterstudy applications. One of the mostpopular case studies is simulating theradiation process within the heaterelements of a nuclear reactor.

At the technical level, there are manynovel technological elements introducedin the ClusterGrid. The first innovativefeature compared to traditional gridsystems is that the labs are connected toone another through private computernetworking (see Figure), using the capa-bilities of the high-bandwidth HungarianAcademic Network in order to enhancesecurity and user confidence in thewhole system.

The second innovative element is the useof dynamic user mapping during jobexecution. One of the most serious

causes of bottlenecks in contemporarysolutions is the insufficient separation ofuser and job credentials, which yieldsmonitoring and authorization problems.Furthermore, it is not necessary in theClusterGrid architecture to have thesubmitter’s user credentials configuredat all clusters or supercomputers in thegrid. This gives much more freedom tojobs traversing through different

resources (in fact the job becomes anatomic unit on which different opera-tions, such as execution, transfer, storeetc can be defined).

The third innovative idea is the Web-service transaction-based, state-full, anddistributed resource broker that providesinteroperable gateway functionality tothose grid systems built on classicaldisciplines using XML/SOAP interface.The broker itself contains simple imple-mentation of all basic grid services, suchas grid information systems, job execu-tion systems, and file transfer subsys-tems.

The fourth innovative element is the jobdefinition format that allows a job to bedefined in both static and in dynamicterms. Jobs are built up in directory hier-

archies, ie all pieces of binary, library,input and output files are encapsulatedinto the structure, and at the same timethe job is also a temporal entity, ie a setof operating system processes ondifferent hosts and the relationships(communication, data transfer) betweenthem. The runtime execution structureprovides the following features: the jobis allowed to take its complete environ-ment to the place of the execution (evena licence file, or organization/virtualorganization certificate), the job can becustomized, and workflow definitionscan be treated as part of the job. Sub-jobs (ie jobs within the master job) canbe defined and executed, and meta-jobs,such as code compilation, can easily betreated as ordinary grid jobs.

The ClusterGrid Infrastructure currentlyinvolving 1100 compute nodes has acumulative performance of 500Gflops/sec (50 billion floating-pointoperations per second), which is compa-rable with that of the top five hundredclusters. The system is also cost effec-tive: the measured performance isachieved at an annual operational cost of40 000 Euro. The framework works notonly for the integration of PC clusters,but also of heterogeneous resources suchas supercomputers.

In the future, NIIF/HUNGARNET plansto improve the national production gridin both qualitative and quantitativeterms. This means improving thenumber of compute nodes to two thou-sand (or more), installing storage nodesand eliminating data network bottle-necks.The introduction of new technicalsolutions such as job gateways, SOAP-interfaced Web portals and experimentalIPv6 grid technologies is also of keyimportance for the forthcoming develop-ment.

Link: http://www.clustergrid.iif.hu

Please contact:Péter Stefán, Office for National Information and Infrastructure Development (NIIF/HUNGARNET), HungaryE-mail: stefá[email protected]

Connection of grid clusters via MPLS VPN over the Hungarian Academic

Network.



Computing clusters can be found in lotsof companies and institutions today.Researchers and engineers have highdata-processing needs and use them todistribute large jobs to a set of homoge-neous machines. The Linux operatingsystem is a de facto standard for clustermanagement, providing easy administra-tion, good performance and a wide soft-ware support basis. Moreover, the set ofavailable libraries and tools makes Linuxa good choice for scientific applications.

Past projects have been focusing on howto aggregate user workstations from theenterprise network to the clusters. In thisway, the company can take advantage ofthe availability of the processing powerof user PCs. However, the real worldshows that most corporate users arerunning Microsoft Windows™, makingit difficult to aggregate user machines tothe corporate cluster, which is based on

Linux. Other approaches such asSETI@home or XtremWeb use acentralized distribution of computingtasks to Internet machines, but do notoffer the smoothness and ease of use of aLinux cluster.

Icatis is developing the ComputeMode™software suite, which smoothly handlesthis specific issue and aggregates usermachines to the corporate cluster. Aserver is installed on the customerpremises and keeps track of user PCsrunning Windows. During cluster usagepeaks, a number of idle user machinescan be aggregated to the cluster. This isdone through a transparent switch of thePC to a secondary, protected mode fromwhich it proceeds into a network bootfrom the ComputeMode™ server, takingadvantage of the PXE protocol. Thispatented technology provides severalbenefits, such as the full isolation of the

PCs’ hard disks: these are not accessiblewhile the PCs are dedicated to clustercomputing. The OS and system configu-ration of a computing PC are also thesame as a PC from the cluster, henceproviding homogeneity and easingadministration.

The system is designed to be very trans-parent to PC owners, and the machinesare only used at times when the PC islikely to be idle (nights, weekends andduring business trips or vacation). If theuser returns unexpectedly while his/herPC is doing a computation, the user canclaim the PC back, and it restores to thestate in which the user left it within oneminute. This includes the user session,open files, on-going programs and thedesktop.

On the administration side,ComputeMode™ offers a Web adminis-tration interface to register/unregistermachines to the system, handle specificsystem image parameters and the usageschedules for the machines (this can bedone automatically), and check usagelogs. The Job Management System(JMS) administration for the clustershows additional machines in thecomputing pool, and priorities can beadjusted using the standard JMS config-uration tools.

Users of the cluster extended throughComputeMode™ do not see muchdifference when ComputeMode™ isinstalled. Job management is done in thestandard way through the JMS. The onlynoticeable difference is the boost in reac-tivity that can be expected when thecluster is heavily loaded. In such casesthe PCs that ComputeMode™ aggre-gates to the cluster provide some extra

by Bruno Richard and Philippe Augerat

Modern research and industry projects require extensive computational power,which is conveniently provided by PC clusters. However, typical use of theseclusters is irregular and shows some usage peaks. Researchers at Icatis aredeveloping the ComputeMode™ software, which takes advantage of idle PCs onthe corporate network and aggregates them to the cluster during usage peaks inorder to reduce its load.

Effective Aggregation of Idle ComputingResources for Cluster Computing

A ComputeMode screen shot.



computational power and processingoccurs faster for the user.

Icatis is a young company, having beencreated in January 2004 after severalyears of investigation and refinement ofits offer. It was successful on thecommercial side: a contract has alreadybeen signed with a major oil and gascompany and in June 2004 Icatis waselected a laureate in the ‘Innovation-Development’ national contest and wona prize from the French Agency forInnovation (http://www.anvar.fr/).

Most Icatis researchers have beenworking within the ID-IMAGLaboratory (http://www-id.imag.fr/), theApache project run by INRIA, CNRS,IMAG, and UJF. ID is a French publicresearch laboratory, which for the pasttwenty years has been researchingconcepts, algorithms and tools for high-performance, parallel and distributed

computing. Successful experimentsinclude the development of a supercom-puter from standard hardware compo-nents such as those that might be foundin a typical large company. An unusualsupercomputer built from 225 standardPCs (733 MHz, 256 MB RAM, Ethernet100) entered the TOP500 contest(http://www.top500.org/) in May 2001and was ranked 385th worldwide forsupercomputing. Other successful exper-iments such as I-Cluster and NFSp, aswell as the Ka-Tools developed in closepartnership with Mandrakesoft(http://www.mandrakesoft.com/) for theMandrakeClustering product (CLICproject from ID-IMAG), have built somesound technical bases for Icatis.

Icatis benefits from a strong experiencein cluster computing and the LinuxSystem, and has strong links with thehigh-performance computing commu-nity.

Some Icatis customers have alreadyevaluated a ComputeMode™ prototypeon their own premises. It has showngood results for peak usage absorptionon an application for seismic terrainexploration, with each job runningseveral hundreds or thousands of tasks.The full product will be released fromQA in December 2004. Among otherfeatures, future versions ofComputeMode will have wider gridcapabilities, such as inter-Cluster loadbalancing, and multiple administrationroles/domains. At the same time, Icatis isworking on a high-end visualizationcluster.

Link:http://www.icatis.com/

Please contact:Philippe Augerat, Icatis CEO, Grenoble, FranceTel: +33 6 76 27 27 92E-mail: [email protected]

In recent years, teaching and learningpractices have been mainly based on theinformation transfer paradigm. Thisfocuses on content, and on the teacher asthe key authoritative figure whoprovides information. Teachers’ effortshave generally been devoted to findingthe best ways of presenting content inorder to transmit information to learners.Unfortunately the current generation of‘e-Learning solutions’ has adopted therather narrow pedagogic paradigm of‘information transfer’. This occurredsimply because it is an easy way to usethe Web’s basic facilities. Failures, suchas massive drop-out rates, are usuallyexplained by a lack of staff awareness inthe use of the Web, rather than criticalreflection on the limits of this approach.

So, the question remains – how do weprovide better access while maintainingor improving quality of learning throughthe use of ICT? The aim of ELeGI(European Learning Grid Infrastructure),an EU-funded Integrated Project (23partners from nine EU countries), is topromote effective human learning bydemonstrating and supporting a learningparadigm shift from the current idea ofinformation transfer to one that focuseson the learner and on the construction ofknowledge using experiential andcollaborative learning approaches in acontextualized, personalised and ubiqui-tous way.

In our vision, the learner has an activeand central role in the learning process.Learning activities are aimed at aiding

the construction of knowledge and skillsin the learner, rather than the memorisa-tion of information. In keeping thelearner at the centre of the learningprocess, personalisation (exploiting theexisting learner’s capability and skills)and individualisation (creating andadapting learning paths according tolearner’s preferences) become relevantaspects to be supported by technologiesthrough the creation of the right context.

The Learning GridThis new vision has two strong implica-tions with respect to technology: first,teaching and learning will move towardthe form of services, and second, thecreation of virtual communities thatprovide services (eg contents, tutoring,searching for learners for sharing experi-

by Pierluigi Ritrovato and Matteo Gaeta

A semantic Grid for human learning is the vision behind the European ELeGIIntegrated Project for the implementation of future learning scenarios based onubiquitous, collaborative, experiential-based and contextualized learning throughthe design, implementation and validation of the Learning Grid.

The European Learning Grid InfrastructureIntegrated Project



ences etc) will be encouraged.Technology must be selected accordingto these implications. ELeGI will adopt aservice-oriented model, which is deeplyintertwined with the use of semantictagging and aligns itself with the globalcommunity, helping to develop trendssuch as OGSA (Open Grid ServicesArchitecture), WSRF and the NextGeneration Grid.

An open distributed service model isbased on a concept of service (which inour context differs from a product) as akind of predefined combination ofprocesses yielding some result (the goalof the service) from distributed, hetero-geneous, asynchronously communi-cating and available resources. The next

generation of Grid solutions will increas-ingly adopt the service-oriented modelfor exploiting commodity technologies.The goal of this model is to enable andfacilitate the transformation of informa-tion into knowledge, by humans as wellas – progressively – by software agents,providing the electronic underpinningfor a global society in business, govern-ment, research, science, education andentertainment (semantic aspects). Werefer to these efforts as the ‘semanticgrid’.

According to this technology scenario,the research at ELeGI should address the

creation of a semantic grid for humanlearning: the Learning Grid.

The Learning Grid is a semanticallyenriched grid that, bringing together thefeatures of Grid and semantic tech-nology, represents our solution for thecreation of future learning scenarios. It isbased on the OGSA model, so it inheritsall the features of that architecture.

As an example of the potential offeredby the Learning Grid, we provide a shortdescription of how it should support allthe phases of an IMS-Learning Designspecification.

Authoring tools for the production oflearning scenarios can rely on know-

ledge-based decision-making systems tosuggest pedagogical models and/oractivities to suit the situation. This isdone using knowledge of both thecontext and the people involved in thescenario (eg starting skills, personalprofiles etc). Furthermore, experts canuse the collaborative features of the Gridto cooperate in modelling the scenario.In this way, the Learning Grid supportsthe analysis, modelling and developmentphases of Learning Design documents.

In the delivery phase, the LearningDesign document should be understoodand its content executed.

The Learning Grid uses its knowledge(represented through OWL-S ontology)to cross-reference learner preferencesand the pedagogical model with thetools, resources and activities availableon the Grid.

The Learning Grid makes available alearning scenario virtualized as a'Human Learning Service', with all its'implicit knowledge' (pedagogicalmodel, learning goals, resources andactivities involved, etc…) as a buildingblock for the creation of more complexand interactive learning experiencescomposed of different scenarios andcontexts.

The Methodology and ApproachThe project’s objectives will be accom-plished following a proactive, integratedand iterative approach. Being proactiveis essential due to the novelty of the tech-nologies that will be adopted. A proac-tive approach will also be adopted forrisk management. The Figure shows theELeGI activities and their relationships.We have selected a set of demonstratorsand testbeds representing scientific,social, economic and cultural cognateareas that include both formal andinformal learning scenarios. The keydifference between testbeds and demon-strators is that demonstrators alreadyexist in non-Grid-compliant forms,whereas testbeds are principally newdepartures, designed to test the ELeGIapproach from conception to implemen-tation and evaluation.

Links:EleGI Web site:http://www.elegi.org

LeGE-WG Web site:http://www.lege-wg.org

IMS-LD specifications: http://www.imsglobal.org/learningdesign/

Please contact:Pierluigi RitrovatoCentro di Ricerca in Matematica Pura ed Applicata, Università di Salerno, ItalyTel: +39 089 964289E-mail: [email protected]

ELeGI activities.



Research work today is often a collabo-rative effort carried out by groups ofindividuals belonging to different orga-nizations remotely spread worldwide.Motivated by a common goal andfunding opportunities, these groupsdynamically aggregate into virtual orga-nizations (VOs) that share theirresources eg knowledge, experimentalresults, instruments, etc., for the durationof their collaboration, creating a rich andpowerful research environment.

The DILIGENT project aims atsupporting this new research operationalmode by creating an advanced know-ledge infrastructure that will serve theneeds of dynamic virtual organizations.This infrastructure will allow membersof VOs to access shared knowledge,services and computational resources ina secure, coordinated, and cost-effectiveway.

DILIGENT will be built by integratingGrid and Digital Library (DL) technolo-gies. The merging of these two differenttechnologies will result in an innovativelevel of functionality providing the foun-dations for next generation collaborationenvironments able to serve manydifferent research and industrial applica-tions.

In particular, the Grid framework willprovide the context for implementing thenotion of virtual digital libraries (VDLs),ie transient, on-demand DLs based onshared computational, multimedia andmulti-type resources. The DILIGENTinfrastructure will maintain a network ofexisting resources on the Grid. A virtual

organization will be enabled to dynami-cally create and modify its own DL byspecifying a number of requirements onthe information space (eg publishinginstitutions, content domain, documenttype, level of replication) and on theservices (eg service type, configuration,lifetime, availability, response time). Areliable and secure virtual DL that satis-fies the given requirements will be trans-parently instantiated and made acces-sible to authorized users through aportal. Many virtual DLs, servingdifferent user communities, will beactive on the same shared Grid resourcesat the same time. The composition of aDL will be dynamic since it will dependon the currently available and registeredDL resources and on many other qualityparameters such as usage workload,connectivity, etc. This developmentmodel will make it possible to avoidheavy investments, long delays andradical changes in the organizationssetting up these applications, thusfostering a broader use of DLs as meansfor communication and collaboration.

The Grid framework will also enable theprovision of a number of new functionswhose implementation has until nowbeen limited by their high cost in termsof computational, storage and datatransfer capacity, such as multimediadocument and geographic informationprocessing, 3D handling, spatial datamanipulation, etc.

From the technical point of view,DILGENT will exploit the results of theEGEE (Enabling Grids for E-science inEurope (http://public.eu-egee.org/)

project, which, in the next two years, willdeliver a Grid production infrastructureshared by a very large number ofEuropean organizations. DILIGENTwill first enrich this infrastructure withthe necessary features for creating andhandling an open and 'on-the-fly' modifi-able set of services as required by the DLapplication framework. It will then add anumber of specific services for: • providing access to content sources,

such as archives, metadata repositoriesand databases

• implementing typical DL functions,like search, annotation, personaliza-tion, document visualization

• supporting existing applicationsimplemented by third-parties, likevideo summarization, reports genera-tion, media-specific similarity search,etc.

The DILIGENT test-bed will be demon-strated and validated by two comple-mentary real-life application scenarios:one from the environmental e-Sciencedomain and the other from the culturalheritage domain. The user community ofthe first scenario is composed by repre-sentatives of leading organizations thatoperate in the environmental sector.DILIGENT will be used experimentallyas a means for supporting two of thetypical activities of this community: theorganization of conferences and thepreparation of projects and periodicalreports on specific environmental topicsof concern. We expect that DILIGENTwill facilitate the achievement of thegoals of these activities and will enhancethe quality of their results by improvingaccessibility, interoperability and

by Donatella Castelli

Building on the results of its past activity in both the Grid and Digital Librarydomains, ERCIM has provided the framework for the setting up of an innovativeIST 6FP funded three-year Integrated Project, DILIGENT. This project, which iscoordinated scientifically by ISTI-CNR, involves 14 European partners and anumber of international observers. These research institutions and companieswill work together on the development of a digital library infrastructure, basedon Grid-enabled technology, that will allow members of dynamic virtualorganizations to collaborate by exploiting shared knowledge and physicalresources.

DILIGENT: A Digital Library Infrastructureon Grid Enabled Technology



usability of environmental data, models,tools, algorithms and instruments and byintegrating the located data sources withspecialized data handling services.

The user community of the secondscenario consists of scholars, distributedall over the world, who have decided towork in a three year research project inorder to set up the basis for a newresearch discipline that merges togetherexperiences from the medical, humanity,social science, and communicationresearch areas. DILIGENT will beexperimented as a means to providethese scholars with a cost-effectiveinstrument for setting up DLs servingtheir needs despite the limited durationand funding resources available for their

project. The system will also be usedexperimentally for the organization ofcourses. Specific VDLs that address theknowledge needs of the students will becreated by re-using material maintainedin the registered archive resources of theDILIGENT infrastructure.

The DILIGENT consortium expects thatthe Grid-based DL infrastructure will beadopted and extended by several othercommunities in order to serve differentapplication areas. Some of these commu-nities have already been registered asproject observers. Regular trainingsessions and workshops will beconducted to disseminate the experienceand results of the DILIGENT project to

both scientists and potential DILIGENTusers.

Information on how to become a DILI-GENT observer and about the DILI-GENT activities and events can be foundon the project website.

Links: http://www.diligentproject.orghttp://public.eu-egee.org/

Please contact:Donatella Castelli, ISTI-CNRScientific CoordinatorTel: +39 050 3152902E-mail: [email protected]

Jessica Michel, ERCIM officeAdministrative CoordinatorTel: +33 4 92 38 50 89E-mail: [email protected]

Business models in Grid computingdealing in buying and selling resourcesacross budget boundaries (within orbetween organizations) are in their veryearly stages. Cycle-harvesting (or -scav-enging, or -stealing) is a significant areaof Grid and cluster computing with soft-ware available from several vendors.However, creating commercial contractsbased on resources made available bycycle-harvesting is a significant chal-lenge for two reasons. Firstly, the char-acteristics of the harvested resources areinherently stochastic. Secondly, in acommercial environment, purchaserscan expect the sellers of such contracts tooptimize against the quality of service(QoS) definitions provided. These chal-lenges have been successfully met inconventional commodities, eg, RandomLength Lumber (RLL), traded on finan-cial exchanges (the Chicago Mercantile(CME, www.cme.com) in this case). Theessential point for creating a commer-cially valuable QoS definition is to guar-

antee a set of statistical parameters ofeach and every contract instance.

The chance of achieving a good QoSlevel of the delivered resource slots justby taking whatever comes and passing iton (ie, basically random sampling of theunderlying harvested resources) isshown in Figure 1. The error levelpermitted is taken as 1% of the contractsize. As expected, larger contracts (moresamples per quantile) make it easier toattain the fixed QoS requirement.However the probability of satisfying thequality requirements with randomsampling is basically zero. Our analysisis valid when quantiles are guaranteedfor any distribution: the analysis has nodependence on the underlying distribu-tion. Thus we need a systematic way todeliver HSQ – we do this by combiningstochastic (harvested) resources withdirectly controlled (deterministic)resources.

Stochastic QoS is a QoS parameter or setof QoS parameters that is based on statis-tical properties of some QoS metric.Hard Stochastic QoS (HSQ) is simply aQoS parameter or set of QoS parametersthat is based on statistical properties ofsome QoS metric and is guaranteed withcertainty. These metrics are guaranteedfor each and every contract. Statisticallythis is the difference between guaran-teeing the properties of a sample versusguaranteeing the properties of the popu-lation from which the sample was drawn.In business terms this means that eachcontract can have a consistent value –not just a collection of contracts or aparticular provider. A typical example inconventional commodities is the specifi-cation of Random Length Lumbercontracts on the Chicago MercantileExchange; in the IT world statisticalguarantees on network usage andcapacity parameters are commonplace.

Cycle-harvesting – locating unused on-line computational capacity and makingit available – is an important feature of Grid resource management. Work at IBMResearch has led to a quality of service (QoS) definition, called Hard StatisticalQoS (HSQ), and an accompanying system architecture that can supportcommercial service contracts.

The Cost of Quality for Cycle-Harvesting Gridsby Chris Kenyon



We have defined a system architecturefor delivering HSQ. This comprises fivethings: an HSQ controller; a pool ofstochastic (harvested) resources; a poolof deterministic (dedicated) resources;monitoring of the stochastic pool ofresources; and control of the stochasticresource pool. The basic idea is that theHSQ Controller monitors the CycleHarvesting System (CHS) resources,using the CHS Monitor, as applied toeach Contract and takes one or more oftwo possible actions: sends an artificially'harvested resource end' to a particularCHS resource; or diverts contract obliga-tions from the CHS resources to theDeterministic Resources, under controlof the Deterministic Controller, withappropriate instructions for their execu-tion.

Given that resource contracts are beingtraded and the owner of the HSQ systemsees an Ask (ie, a request for resources)with a price and a deadline, what is thecost to support that request? If it is belowthe price then the HSQ system owner canrespond directly to the Ask. If not thenthe owner can post an alternative price.The cost will consist of the quantity ofstochastic resources required and therelative proportion of deterministicresources to meet the quality require-ments together with the costs of the twotypes of resources. In many situationsthe major cost will be the deterministicresources.

The motivation for this work was thedesire to see whether it was possible toturn cycle-harvested resources intosomething commercially valuable. Wasit possible to define a contract that wasvaluable given that the underlyingresources are inherently stochastic andthat in a commercial environment we canexpect the provider of a QoS guaranteeto optimize against it? We wanted toavoid all arguments based on long termreputation because we were interested inhaving valuable (tradable) individualcontracts for cycle harvested resources –not the question of how to create valu-able reputations for the providers.

We have shown how HSQ can be imple-mented for cycle-harvested resourcesusing a hybrid stochastic-deterministicsystem. We have also described architec-ture, and algorithms, to support HSQcontracts. We have analyzed the algo-rithm behavior analytically using adistribution-free approach in terms of theexpected proportion of deterministicresources required to support a givenHSQ level. Thus our analysis can beapplied whatever the details of a partic-ular system happen to be.

For a particular HSQ example (seeFigure 2) where time slot lengths werelog-Normally distributed we found thatto provide hard guarantees on 8 quantilesfor contract sizes of 16 to 1024 slots,from 13% to 1% additional deterministicresources were required. Permittingoversampling, for example for a contract

size of 64, was relatively inefficient,leading to up to 61% of the stochasticresources being wasted. Including down-wards substitution (ie, accepting longertime slots as valid substitutes for shorterones) reduced deterministic resourcerequirements by roughly half. Theseresults are robust (change <5%) forchanges in shape and scale parametersby a factor of 2 in both directions.

We have concluded that commercialservice contracts with hard statisticalQoS guarantees (HSQ), based on cycle-harvested resources are viable both froma conceptual point of view and quantita-tively with only small (1% – 10%)requirements for deterministic (dedi-cated) resources.

Links:http://www.zurich.ibm.com/grideconomics

Please contact:Chris Kenyon, IBM Research, Zurich Research Lab., SwitzerlandE-mail: [email protected]

Figure 1: Probability of fulfilling a contract (Prob[success]) for a

given distribution of slots using random sampling from the

harvested resources relative to the precision with which the

distribution is defined (number of quantiles). These results are

valid for any distribution.

Figure 2: Expected proportion of deterministic resources required

(E[R]) to guarantee a given distribution (LogNormal with mean 3

and variance 1.5) to different levels of precision as defined by the

distribution’s quantiles.


29

The aim of the JGrid project is todevelop a software framework for next-generation Grid systems. We believe thatfuture grids will be pervasive, enclosingvarious software and hardware compo-nents as services, and will providesimple and seamless access to servicesfrom anywhere at any time. Future gridswill also be used not only for high-performance computing tasks, but alsofor carrying out business and assistingsociety and individuals in many aspectsof everyday life. These systems willneed to be reliable, to operate withminimal administrator supervision; self-organizing, adaptive and self-healingbehaviour is mandatory.

The project is being pursued in Hungaryby a consortium of two universities(University of Veszprém and EötvösUniversity of Sciences, Budapest), oneresearch institute (SZTAKI) and SunMicrosystems Hungary. It is supportedby the Hungarian Government underIKTA grant No. 089/2002. The project isalso listed as a Jini community project,meaning it is open to the internationalresearch and development community(see the links section for addresses).Work started in 2002 and is expected tofinish at the end of this year.

Among the most challenging problemsposed by grid systems are discovery andcoordination, that is, how to find andorchestrate a set of services operating onheterogeneous resources under differentadministrative control in order to solve agiven problem. Grids are dynamic;services may join and leave the system atany time by their own will or due tofaults. Current grid frameworks do notadequately support dynamic servicediscovery. Heterogeneity is addressed bycreating a common (XML-based)protocol to achieve interoperability.Next-generation grid applications thatuse a large set of services need more

however; a programming model isrequired that enables the creation ofapplications that can execute in suchhighly dynamic environments. Thismodel needs to support, among otherthings, state-full service invocation,error handling, notification mechanismbased on events and security.

The Java language is a good candidatefor creating grid systems and a grid

programming model. It is an object-oriented, platform-independent languagewith built-in security and error handlingfeatures. On its own, however, it isunsuitable for creating dynamic systems.Jini Technology – developed by SunMicrosystems in 1999 – provides themissing features for the Java platform.Jini is an infrastructure and program-ming model for creating dynamicdistributed systems. It supports service

The JGrid system is a service-oriented grid framework that enables theconstruction of dynamic, self-organising grid environments that support globalservice discovery and provide seamless, secure access to a wide range of gridservices on demand.

JGrid: A Jini Technology-Based Service Gridby Zoltán Juhász and Gergely Sipos

Figure 1: Part of the JGrid Service

Browser graphical user interface

showing discovered services.

Figure 2: Overview of the job execution process in the JGrid system using a broker. The

client contacts the broker and specifies the type of services sought for job execution.

The broker discovers available services S1-S9 and selects the most suitable one, S4. It

then submits the job to this service. Status events are sent to the broker and then to the

client to monitor execution. The broker may choose alternative services in case of a

service fault.




discovery, protocol-independent serviceaccess, event-based notification,distributed resource management andsecurity, and hence is a natural fit forcreating grid systems.

Currently, the main focus of the JGridproject is on computational problems;we have developed key grid servicesrequired for the construction of compu-tational grids. JGrid users can createexecute-interactive grid applicationsusing Java tasks that execute onCompute services. Tasks can be sequen-tial or parallel. Parallel programs can useMPI-style message passing or remotemethod invocations for inter-processcommunication. Non-Java applicationssuch as legacy Fortran and C programscan be executed on Batch services thatintegrate batch runtime systems (eg SunGrid Engine or Condor) into the JGridframework. User data can be stored inthe grid using services that allow users tostore and retrieve their files and datafrom anywhere in the world. JGridprovides wide-area discovery thatextends the Jini discovery mechanism inorder to efficiently find grid services onthe Internet using powerful search

queries. A service browser is provided tofind and display available services to theuser. Discovering and controllingservices may be a tedious task. The JGridCompute Broker service relieves usersfrom system-related tasks; a job can besubmitted directly to the broker withappropriate resource requirements, andthe broker will then discover suitablebatch services, submit the job and super-vise its execution. On detecting an error,it may take corrective measures onbehalf of the user such as re-submittingthe job to another batch service; onsuccessful completion, the user receivesthe result. The broker provides computa-tional users with a largely simplifiedinteraction with the grid.

The JGrid project also provides develop-ment support for grid applicationprogrammers. The P-Grade graphicalprogram development environment hasbeen extended in order to generate JGridexecutables from graphically designedparallel message passing programs. TheP-Grade system also supports thecreation of grid workflow applicationson JGrid. The graphical workflow editorof the P-Grade portal system can be used

to define a workflow, which is thenexecuted under the supervision of aJGrid Workflow Manager service.

The JGrid system has been demonstratedon several international forums and canbe downloaded from the project home-page. A national testbed is underconstruction that will allow users toexperiment with JGrid services, developand execute grid applications. Futureproject activities include the develop-ment of non-computational services, egfor media streaming and processing,database access and e-commerce andbusiness applications, as well asproviding access to services from mobiledevices.

Links: http://pds.irt.vein.hu/jgridhttp://jgrid.jini.orghttp://www.sun.com/jini http://www.lpds.sztaki.hu/pgrade

Please contact:Zoltán Juhász, University of Veszprém, HungaryE-mail: [email protected]

Application service providers empoweredby GRASP middleware are no longerproviding a fixed solution to the customerwhere the business logic is hardwired inthe server side code but large portions ofthe business logic are achieved bycustomising and combining private andoutsourced services as needed.

When properly integrated into a businessworkflow this bundle can soften enter-prise borders, giving way to new, moreflexible ways of secure and reliablecollaboration. To achieve this, GRASPhas developed an architectural frame-work for Grid-based Application ServiceProvision and a prototype realization of

this framework. GRASP integratesconcepts from application service provi-sion and utility computing, from ServiceOriented Architectures and WebServices, and a service-oriented view ofGrid computing in order to empower thisnew generation of grid-based applicationservice providers.

by Theo Dimitrakos

GRASP is an industry driven European research project exploring the use of theGrid Services concept for providing an effective technological basis supportingthe evolution of the ASP market towards a sustainable Utility Computing model.The intention is to improve enterprise application service provision models soas to be able to take full advantage of the flexibility offered by Web Services andthe Grid for supporting dynamic resource allocation, life-time management ofdynamic service instances, resources integration and efficient distributedcomputation.

The GRASP Project: Innovative Middleware for Next Generation Grid-Based ApplicationService Providers



Grasp Architecture Building on top of existing OGSAcompliant middleware the key function-ality of Location, Instantiation andOrchestration of Grid Services isextended to handle other enterpriseconcepts:• Subsystem "Location" to enable a

dynamic search of Grid Services basedon functional, performance, techno-logical and business related criteria.

• Subsystem "Instantiation" to attend theproperly instantiation of Grid Servicesperforming Load Balancing and invo-cation of other related subsystemssuch as Accounting or SLA moni-toring.

• Subsystem "Orchestration" to supportthe ASP in the provision of new busi-ness services achieved by means of thecomposition and coordination ofexisting Grid Services. It implementsthe workflow that encapsulates thebusiness logic of the ASP application(described as BPEL businessprocesses).

Further business functionality isprovided as part of the GRASPFramework by means of a number ofservices classified in different subsys-tems that address core business supportoperations including:• Subsystem 'SLA management' to

support a SLA management systemincluding the use of SLA templates fornegotiating service provision based onQoS criteria, and to monitor the fulfil-ment of contract associated to each

Grid Service instances regarding QoSwhich are based on technical and busi-ness metrics in the vision of GRASP.

• Subsystem 'Accounting' to perform thecharging of service usage based onservice related contracts betweenservice users and service providers,obviously a must for ASPs. ThisGRASP subsystem works out thecomposite account of all services thatare being executed on behalf of aclient. It takes care of the collection ofraw performance/resource data(shared with SLA monitoring);merging different costing models andapplying explicitly defined pricingalgorithms.

• Subsystem "Security" to offer flexiblesecurity perimeters in which GRASPinfrastructure secures completeApplication Service (multipledistributed component Grid Services).It handles securely life-time and addi-tion and expulsion of component Gridservice instances collectivelyexecuting a complete application.

Overall, GRASP has been designing andimplementing business critical subsys-tems that have not been addressed as yetin other major Grid projects includingEDG, Globus and Unicore.

GRASP Implementation StrategyThe current prototype of the GRASPproject middleware leverages on the useof an 'open standards' technology baseline including W3C web services stan-dards and Microsoft, IBM, and BEA

championed WS-* extensions (notablyBPEL4WS, WS-Security, WS-SecureConversation and WS-Trust) andOGSI reference implementations.Continuing its committeemen topioneering the convergence of WebServices and Grid Technologies for busi-ness oriented Grid applications, theGRASP consortium is currently devel-oping a strategy for migrating its imple-mentation base-line from compliancewith OGSI standards to compliance withWSRF standards. A preliminary assess-ment indicates that because of the waythat GRASP architecture has beendesigned (eg loose-coupling betweenGrid middleware functions and Grid-aware business and generic applicationcomponents) the cost of such a migrationis substantially low and we expect thenext generation of GRASP middlewareto support WSRF concepts.

Links: Project Website http://www.eu-grasp.net

GGF OGSA WG: https://forge.gridforum.org/projects/ogsa-wg

WSRF: http://www.oasis-open.org/committees/wsrf

OGSI.NET: http://www.cs.virginia.edu/~humphrey/GCG/ogsi.net.html

Please contact: Theo Dimitrakos, CCLRC, UKTel: +44 1235 446387E-mail: [email protected]

Key concepts underpinning

the GRASP project vision.



Akogrimo will bring together the marketorientation and pervasiveness of mobilecommunication technology in everydaylife with the promise of a dynamicallyconcerted use of resources and servicesprovided through Grid infrastructures.

By integrating the widely disjoint worldsof data communication, telecommunica-tion and distributed service architectures,the final result of the Akogrimo exercisewill be a commercial operator-orientedarchitecture and platform, whichsupports the deployment of Grid servicesin a worldwide perspective.

Selected scenarios showing the poten-tials of the innovative Akogrimoapproach for the quality of life of thepublic citizen, for business opportunitiesfor both small and medium enterprises aswell as large companies includingnetwork providers will validate theAkogrimo architecture.

The vision of Akogrimo is accordingly aworld in which:• Grid services, pervasively available,

are eventually meeting the ‘every-where at any time in any context’paradigm

• Grid services, comprising personal-ized knowledge and semantics, areallowing for ad-hoc, dynamic, andpossibly federated formation ofcomplex problem solving scenarios ineveryday life, in business and science -to fixed, nomadic and mobile citizens

• network and service operators are ableto develop new business activities andto provide profitable services in suchan integrated world based on Grid andMobile communications concepts.

ScenariosThe realization of a challenging frame-work merging Grid service architectureswith the underlying mobile networktechnology requires an iterative realiza-tion approach. Iterations of the frame-work need one or more non-trivialscenarios that are able to validate thefeatures and capabilities that are added to

the framework. The approach withinAkogrimo foresees to define, in thescenario specification phase, perfor-mance requirements and to validate earlyversions through prototypes or ‘valida-tion scenarios’ testing a limited set of thefunctionality. Following the frame-work’s evolution, the validationprogresses from system component levelto the complete demonstrator.

The three scenarios foreseen so far forvalidation impose demands that firstlyreflect requirements from ‘traditional’Grid applications, namely access todistributed data and compute intensiveservices. In addition the scenariosdemonstrate the need for mobile

dynamic virtual organizations (MDVO)exhibiting the ability to dynamicallyadapt the organizational structure tochanging local situations (throughcontext awareness, availability of sharedmobile resources). Within theseMDVOs, complex workflows basedupon Grid infrastructure and Gridservices to access data from distributed,

sometimes evenmobile databaseswill be to dynami-cally established andprocessed. Of partic-ular relevance is theintegration of loca-tion based servicesand spontaneoususability.

The IP(v6) capablenetwork supportingthe scenarios andtheir related require-ments will beconstructed and

derived from the infrastructure built inStuttgart and Madrid in the context of theIST Moby Dick and Daidalos projects.This infrastructure integrates IP-basedMobility, AAA, QoS and network secu-rity on a pure Mobile IPv6 platform.

At the moment the following scenarioshave been identified:

Akogrimo eLearning The Akogrimo eLearning scenario willbe embedded in the frame of the E-Learning domain. The focus in thisscenario will be to build a showcase fornew ways of learning that is madepossible by the Akogrimo infrastructure.It is planned to liaise with the Integrated

Mobility has become a central aspect of life for European citizens - in business,education, and leisure. Due to rapid technological and societal changes, therehas been an astonishing growth of technologies and services for mobile users.Large investments have been made in order to provide the necessaryinfrastructures across Europe. In 2003 the number of Internet connected handsetsor mobile hosts equalled the number of fixed Internet connected. Taking intoaccount this evolution, Akogrimo — by leveraging the large base of mobile users— is aiming to radically advance the pervasiveness of Grid computing acrossEurope.

Akogrimo — The Grid goes Mobileby Stefan Wesner, Theo Dimitrakos and Keith Jeffrey

The 'Next

Generation

GRIDNET'.



Projects ELeGI and to utilise the MobileUniversity testbed of the Daidalosproject in order to increase the impact.

Akogrimo eHealth The Akogrimo eHealth testbed exploresGrid technology and the Mobilityparadigm in the healthcare domain. Itbuilds on previous successful resultsfrom a 6-year German priority researchprogram. Target users of a Grid basedhealthcare information system are theEuropean citizens, mobile pervasivedemand for healthcare services (egchronic diseases), and from the health-care service suppliers all types of health-care service providing institutions andstationary or mobile professionals whichinclude healthcare advisors, pharmacies,

nursing services, hospitals, emergencyservice devices and emergency stations.

Akogrimo Disaster Handling and Crisis Management (DHCM)Akogrimo DHCM involves incidentswhere various crises or disasters shouldbe handled by rescue services and othermission-critical mobile personnel, whohave to collaborate within time-criticaland dangerous situations such as largesport events, concerts or special loca-tions such as airports or railway stations.

ConclusionAkogrimo adds a new dimension to theGrid. Mobility and network integrationhas not been addressed so far by otherinitiatives. With a merged infrastructureenabling cross-layer communication

between network and Grid middlewarenew kind of applications are possiblethat attract not only researchers inspecific domains but enable the provi-sion of services for the public citizenmaking a pervasive Grid possible.

Links:Akogrimo: http://www.mobilegrids.orgMoby Dick: http://www.ist-mobydick.org/Daidalos: http://www.ist-daidalos.orgELeGI: http://www.elegi.org

Please contact:Antonio Sánchez Esguevillas, Telefónica,Project CoordinatorTel: +34 983 367 577E-mail: [email protected]

Stefan Wesner, High PerformanceComputing Center, HLRS, Stuttgart, Germany, Project ManagerTel: +49711 685 4275E-mail: [email protected]

The Padico project, which commencedin 2000, is investigating different aspectsof software-component programming toallow the design of middleware andapplications for large-scale distributedsystems and grids in particular. Weconcentrated on the challenging problemof designing runtime systems to supportefficient execution of software compo-nents within a grid. We also proposedextending an existing software compo-nent-model to support coarse-grainparallelism within a software componentwhile maintaining scalable connectionsbetween components. Two INRIAresearchers and three PhD students,within the PARIS project-team atIRISA/INRIA, were involved in thisresearch.

The motivation behind this research wasto handle the complexity of grid infras-

tructures by promoting the use of soft-ware components both for the design ofthe grid middleware itself and the designof applications. However, existing soft-ware component models (OMG CCM,Sun EJB, Microsoft DCOM) are unsuit-able for grids since they were notdesigned to handle large-scaledistributed systems, to use variousnetworking technologies or to supportcoarse-grain parallelism. Moreover,automatic deployment of softwarecomponents on available resources wasnot possible. The PARIS project-teamstarted several research projects to dealwith these problems.

One of the first projects was to extend anexisting component model to supportcoarse-grain parallelism. We selectedCORBA from the OMG as the targetmodel simply because it was a standard

with several existing open-source imple-mentations. We first proposed an exten-sion of the distributed object modelproviding distributed parallel CORBAobjects aiming at encapsulating parallelSPMD codes easily and efficiently. Aparallel CORBA object is just a collec-tion of identical standard CORBAobjects managed as a single entity. Datadistribution and communication betweenparallel CORBA objects is being donetransparently to the programmers. Thisextension of the object model imple-mented in PaCO/PaCO++ was laterincorporated in GridCCM, a parallelextension to the component modelproposed by the OMG (CCM). It is thuspossible nowadays to design multi-physics applications based on theconnection of parallel components, eachof them solving a particular physics.

by Christian Pérez

Recent advances made by researchers from the IRISA/INRIA ‘PARIS’ project-teamin the design of runtime systems will allow automatic deployment of softwarecomponents on large-scale distributed systems and grids while making the bestuse of the available resources (computers and networks). The design of next-generation grid middleware is based on the concept of software components anda new approach to implementing complex multi-physics applications.

A Component-Based Software Infrastructurefor Grid Computing



In parallel, we worked on the design of aruntime system called PadicoTM, toallow components to exchange dataindependently of the underlyingnetworking technologies. Most of theruntime systems associated with existingcomponent models are only able tocommunicate through the plain TCP/IPprotocol. Efficient networking technolo-gies found in clusters were thereforeinefficiently used. Similarly, high-band-width Wide Area Networks are under-

exploited without the use of advancedfeatures such as parallel streams. Thanksto the PadicoTM communication frame-work, software components canexchange data in a very efficient waywhatever the networking technologies(SAN, LAN, WAN).

One of the issues under current investi-gation is the automatic deployment ofsoftware components onto grids. Wepropose a network topology description

model that is capable of describing theavailable networking resources and theirconfiguration. This model has beenimplemented in the Globus middlewarewithin the information service. It willallow an efficient mapping of the appli-cation component graph to the gridinfrastructure, taking into considerationthe application requirements.

These different projects are being inte-grated in a single software platformbased on Globus. We are cooperatingwith different project-teams withinINRIA to experiment with real applica-tions. One of them deals with simulationof fluid and solute transport in sub-surface geological layers. It requires thecoupling of several parallel codes forsimulating salt-water intrusion and reac-tive transport. The software componentmodel provides a convenient model forhandling the complexity of the applica-tion while hiding the complexity of thegrid.

This research received financial supportfrom the ACI GRID – the Frenchnational program funding Grid researchactivities. Researchers involved in thePadico project are A. Denis, S. Lacour,C. Pérez, T. Priol and A. Ribes.

Please contact:Christian Pérez, INRIA, FranceTel: +33 2 99 84 72 05E-mail: [email protected]

Figure 2: Example of a complex grid and its network topology description.

Figure 1:

The PadicoTM

framework

decouples the

runtimes from

the actual

networks.

An approach is to organise a communaldata resource so that the data is main-tained in one place, under one organisa-tional scheme. There are three reasonswhy this doesn't work well for researchdata:

• Much of the data has already beencollected under different regimeschosen by the groups who conductedthe work.

• To integrate the collection processitself would lose a valuable asset – theskill and knowledge of the collectors –

they understand their domain andknow best how to organise it for theirwork.

• It is incompatible with human nature –people want to 'own' their own workand decide how it should beconducted; in health care and engi-

We can make decisions and discoveries with data collected within our ownbusiness or research. But we improve decisions or increase the chance and scopeof discoveries when we combine information from multiple sources. Thencorrelations and patterns in the combined data can support new hypotheses,which can be tested and turned into useful knowledge. We see this requirementin medicine, engineering, sciences, economics and business.

Data Access and Integrationby Malcolm Atkinson



neering, I have seen the process oftrying to agree on a common model goon indefinitely, consuming the time ofmany experts, but never converging.

An important trend in research is toorganise larger shared data resources,but at the same time the number ofcollections is growing rapidly. Theirdiversity increases, while multi-disci-plinary research increasingly requiresintegration of data from multipleautonomous and heterogeneous dataresources managed by independentcommunities.

The Data Warehouse ApproachAnother approach is the data warehouse– much used in retail and financial deci-sion support for example. Here, datafrom multiple data resources is copied,'cleaned' and 'integrated' under acommon schema. Typically, more data isadded to the data warehouse from a stan-dard set of sources on a regular periodicbasis. This works well with a small set ofstable data resources serving well-defined goals. In research, there is alarge and growing set of resources, anopen-ended set of goals and each sourcechanges, both in content and structure, asexperiments and surveys progress and asunderstanding and technology advances.Data warehousing has its place insecuring and concentrating data, butencouraging and capturing multi-sourceevolution is vital. It becomes part of thescientific communication processes andenables rapid use in other domains ofnew data, new discoveries and new clas-sifications.

A virtual data warehouse (I am indebtedto Mark Parsons, EPCC, Edinburgh forthis term.) would meet this requirement.It would enable researchers to combinedata from a dynamically varying set ofevolving data sources. It would accom-modate all of the diversity, handlingschema integration (the differences inthe way data is structured and described)and data integration (the differences inthe sets of values used and their repre-sentations). It would accommodate thechanges in policy of data owners, and thechanges in organisation and content ofeach data source. And it would do all ofthis while presenting its users withunderstandable and stable facilities, that

nevertheless reflect the new information.It would provide them with confidenceand provenance information, so that theycould use the resulting data as reliableevidence. At the same time, it would notplace restrictions on the autonomy of thedata providers, would assure them thattheir data was not misused and assist inensuring they gained credit for thequality and content of their data.

A complete realisation of a virtual datawarehouse is beyond the current state ofthe art, instead various aspects of it arerealised. Many research teams hand-craft a one-off solution. This does notscale well, as the skilled work ofbuilding schema and data mappingsgrows with the number and richness ofthe data sources integrated. It is vulner-able to being unsustainable as thoseresources evolve and rarely implementsmechanisms to ensure data access poli-cies and data provenance tracking.Projects, such as myGrid and DiscoveryNet craft workflows to assemble relevantdata for each analysis. The Virtual DataTechnology (Pegasus, Chimera, andDagMan) developed by the GriPhyNproject, encourage high-level expressionof combined data integration and anal-ysis, enabling the underlying system toplan and evaluate more optimally. VDTexploits the predominant update patternin physics data of incremental additionagainst a constant schema. Projects suchas BIRN and GEON at SDSC cataloguedata from multiple sources, describingtheir structure and data representation.For example, each geological survey’srock classification is described with anontology and its coordinate system isdefined. Tools are provided to managethese descriptions and use them toconstruct schema and data mappings.

Technical solutions to recurringdistributed tasks underlie the assemblyof data in data warehouses, communal orshared repositories, bespoke solutions,catalogued registries and in advancestowards virtual data warehouses:• data description and discovery• access, authentication, authorisation &

accounting• data transformation• data transport• query, selection and aggregation• data update, bulk load and archiving

• provenance tracking, audit trails anddiagnostic logging.

The UK’s OGSA-DAI (Open GridServices Architecture Data Access andIntegration) system provides a frame-work and set of components operatingacross grids or web services to deliverthese mechanisms reliably. Using thisframework, many projects are buildingon OGSA-DAI and extending its reper-toire. It already handles a range of rela-tional systems, XML databases andcollections of files.

Perhaps the most important challengesthat remain are:• Automatic adaptation of schema and

data transformations in response tochanges in a data source.

• Integrated optimisation of computa-tion, data management and data move-ment.

• high-level presentation of data compo-sition, operations, analyses and trans-formations.

All three require a formal foundation,based on tractable mathematical modelsof sufficient scope and realism.

Link: UK National e-Science Centre: http://www.nesc.ac.uk/

OGSA-DAI:http://www.ogsadai.org.uk/

Please contact: Malcolm Atkinson, National e-Science Centre, Edinburgh, UKE-mail: [email protected]



It is expected that as the Grid matures,user communities will become moredynamic, shorter-lived, and based ondetailed policies rather than binaryaccess decisions. Policy-based coordina-tion of all resources, and minimal per-resource configuration is thereforeessential.

Currently, most Grid infrastructures useas their core authorization system GSI,the Grid Security Infrastructure devel-oped by the Globus Alliance. GSI isbased around the Public KeyInfrastructure concept, where trustedthird parties issue certificates to bothusers and resources. In plain GSI, accessrights are directly linked to the user'sidentity. The resource provider keeps alist (grid-mapfile) of names of autho-rized users, and associates local accountswith these users. However, it isnecessary to make the mappings inadvance, as the grid-mapfile is main-tained locally at each resource.

Unfortunately this does not scale well,since the grid-mapfile must be updatedcontinuously as users are added to thecommunity or ‘virtual organization’(VO). It is also prone to privacy prob-lems, since it forces one to publish a listof all users in a project to all partners,even if the user never intends to use aspecific resource.

Local Credential MappingThe local authorization services devel-oped by NIKHEF mitigate these prob-lems. The new model is based on theVirtual Organization ManagementService VOMS, an authorization-attribute service developed in theDataTAG project by the Italian nuclearphysics institute INFN. Access controlsare no longer bound only to the user's

identity, and decisions can be based onattributes issued to the user by the VO.The only infrastructure needed to estab-lish a VO is this attribute authority, andat the local resource a single line ofconfiguration suffices both to enableaccess for the new community and tosupport fine-grained policy controlsdefined and implemented by the VO.

Recently, the Virtual Laboratory for e-Science project in the Netherlandsadopted this model for enforcing accesscontrols to data sets spread across twohospitals, a research institute and thenational computer centre.

The site aspect of the system imple-mented by NIKHEF consists of twoparts: the Local Centre AuthorizationService (LCAS) and the Local CredentialMapping Service (LCMAPS).

LCAS is a policy-decision module forimplementing site-access policies, wherethe policies can be changed at run-time.It is currently a key element in enforcingsecurity policies in many productionsystems, enabling site security officers tolimit damage caused by compromisedidentities.

However, access control to the resourceas a whole is not sufficient. LCMAPS,which can parse VOMS attributes, candynamically assign specific Unix groupsto fine-grained attributes. These groupsare then linked to an unprivilegedaccount for execution. The translationfrom VOMS attributes to Unix groups isperformed on the fly. Wildcard VOgroup specifications give the site controlover the number of Unix groups assignedto an individual VO. Sandboxing is thusachieved inside any existing Unixsystem.

LCMAPS supports a variety of ways tocollect VO information from traditionaluser-proxies, VOMS attributes and grid-mapfiles, and a wide variety of account-enforcement modules for both singlemachines and clusters of computers.

Multiple ResourcesComputing and storage are only part ofthe picture however, with new resourcetypes such as remote instruments andoptical networks joining the grid. Thisposes new challenges for resourcebrokers when independent organizationsare involved.

Wide-area networks in particular havemany different stakeholders. Packetstravelling from NIKHEF in Amsterdamto, say, the Laurence Berkeley lab in SanFrancisco will encounter five differentproviders. With valuable resources such

Joining resources from different administrative domains is currently a nightmare,but the problem may be alleviated with local and global authorization systemsdeveloped by the Dutch high-energy physics institute NIKHEF and the Universityof Amsterdam (UvA). By combining generic authentication, authorization andaccounting (AAA) service techniques and site-local authorization, the creationof a ‘virtual organisation’ can become more dynamic and easier for both usersand administrators.

Building a Multi-Domain Gridby Freek Dijkstra and David Groep

Apart from computing services, users

may want to provision dedicated network

services at the same time.


37

as optical networks - your own personallight path from A to B - access controland dynamic authorization are sine quanon in the commercial world.Consequently the user running a calcula-tion in Amsterdam, with the data in SanFrancisco, will need both of theseresources simultaneously.

The AAA server, developed by theUniversity of Amsterdam, fills this gapin traditional resource management. TheAAA server is capable of taking autho-rization decisions across domains, anduses policy files to determine businesslogic. It can, based on the current policy,

contact other services to fulfill users’requests. The attributes themselves mayin fact be issued by the VOMS server,thereby seamlessly integrating network,storage and computer access.

The University of Amsterdam (UvA) hasalready demonstrated that the AAAserver is able to make authorization deci-sions involving the provisioning of adedicated network connection, even ifthe connection crosses multipledomains.

Now, the UvA and NIKHEF have joinedforces to build a role-based authorization

system for the European GridInfrastructure, deployed by two projects:Enabling Grids for E-science in Europe(EGEE) and the Dutch VirtualLaboratory for e-Science (VL-E).

Links:http://www.nikhef.nl/grid/http://www.science.uva.nl/research/air/

Please contact:Freek Dijkstra, Universiteit van Amsterdam, The NetherlandsTel: +31 20 5257531E-mail: [email protected]

David Groep, NIKHEF, The NetherlandsTel: +31 20 592 2179E-mail: [email protected]

Emerging new technology for opticalnetworks will deliver QoS and band-width far beyond the capacity and capa-bilities of today’s networks. For the eval-uation of new network components andarchitecture, the integration of networktechniques and applications has proven asuccess in former testbeds. The technicalbasis of the testbed is comprised ofoptical network components connectingcompute resources, from which a gridbased on the UNICORE system is built

up. To allow large complex applicationsstressing the capabilities of the under-lying optical network, additional gridcomponents will be adopted, including aMetaScheduler (allowing co-allocationof compute resources, network resourceswith the necessary QoS or otherresources like visualization devices) andMetaMPICH communication librariesfor distributed applications using MPIfor interprocess communication.

The expected outcomes of the project arethreefold. First, new network techniqueswill be deployed and tested in an opticaltestbed, and along with ambitious appli-cations will provide know-how forfuture generations of networks, espe-cially the next generation of the GermanNREN X-WIN. Second, the enhancedGrid middleware originating from theproject will become useful in theGerman e-Science Initiative D-Grid. Thethird important aspect of the project isthe collaboration with other projects on anational, European and internationallevel.

The consortium of the project is led bythe DFN, ranges from research institutesand universities to the telecommunica-tion industry, and includes ResearchCentre Jülich, Fraunhofer Institute forScientific Computing and Algorithms(SCAI), Fraunhofer Institute for MediaCommunication (IMK), the Centre ofAdvanced European Studies andResearch (CAESAR), RWTH AachenUniversity, Bonn University, theUniversity of Applied Sciences Bonn-

by Helmut Grund and Wolfgang Ziegler

The German Ministry for Research and Education (BMBF) launched the VerticallyIntegrated Optical Testbed for Large Applications (VIOLA) project in spring 2004.The project is managed by ‘Deutsches Forschungsnetz’ (DFN) and will run forthree years.

Resource Management in an Optical Grid Testbed

Figure 1:

VIOLA

Testbed

Topology.

10 Gbit/s SDH10 GE2.5 Gbit´s (2xGE)planned in phase 2optional in the context of collaboration of GSN+ and VIOLA

Universityof Bonn

Location X

Fraunhofer,Sankt Augustin

caesar, Bonn

FH Rhein-Sieg,Sankt Augustin




Rhine-Sieg (FHRS), Alcatel SEL AG,Siemens AG and T-SystemsInternational GmbH.

VIOLA-Support for Set-Up and Operation of the GridThe initial testbed providing 10 Gbit/sconnection between the sites is located inthe region of Cologne-Bonn-Aachen,with a connection to GEANT/GRANDEin Frankfurt and an extension to Bavaria(see Figure 1, Testbed Topology). Allnodes of the testbed clusters may inter-connect with their Gbit interfacesthrough the underlying optical testbednetwork. The dynamic reservation andallocation of bandwidth upon userrequest makes the network technology asubstantial building block of evolvinggrids. Nevertheless today’s middlewarelacks the functionality to handle this andmake it available to the end user andhis/her applications. Enhancing existinggrid middleware ‘UNICORE’ to create agrid infrastructure based on opticalnetwork technologies is a substantialtask in VIOLA. Research Centre Jülich,Fraunhofer Institute SCAI and theUniversity of Applied Sciences Bonn-Rhine-Sieg are responsible for this task.

Interaction with network resources isneeded on two levels. The first isscheduling of network resources throughdedicated (local) resource managementsystems in a similar way to compute

resources today, whereas the resourcemanagement system for (optical)switches of the testbed implements thenecessary protocol. The implementationof the resource management system isdone by Fraunhofer IMK and BonnUniversity. The second is co-allocationof network and compute resources for agiven task of the user. This will be donethrough implementation of WS-Agreement and WS-Negotiation asproposed by the Global Grid Forumworking group GRAAP (Grid ResourceAllocation Agreement Protocol).

Two requirements for local schedulingsystems must be satisfied: local systemsmust handle the negotiation protocol,and must be able to do advance reserva-tion of resources for a negotiated time-slot. These requirements also guaranteemaximum local site autonomy, since thenegotiation respects the local schedulingmechanisms and site policies remainfully effective. User authentication andmapping is handled by the localUNICORE system as usual.

Current State of the ProjectA Web-service-based version of theMetaScheduler is now ready and co-allo-cation has been tested locally. This firstversion interacts with the UNICOREclient, receiving the job requirementsand sending back the results of the nego-tiations with the local scheduling

systems, ie simple agreements. TheUNICORE system then takes responsi-bility for the user’s job. At the time thisarticle will appear, the necessary modifi-cations of the current version ofUNICORE will be almost complete, andevaluation of co-allocation acrossmultiple sites is expected to start by theend of 2004 (see Figure 2,MetaScheduler Integration inUNICORE). By the end of 2004, a stableversion of MetaMPICH will also beavailable and initial tests running appli-cations in the VIOLA middleware envi-ronment will take place in early 2005.Once MetaMPICH becomes available,parallel I/O will be implemented on topof it. The basic optical network connec-tivity will also be operable by the end of2004.

Future PlansIn the last third of the project, a substan-tially revised UNICORE version basedon OGSA and WSRF will be made avail-able as a middleware layer for theVIOLA testbed. The MetaSchedulingservice as mentioned above will interactdirectly with the new UNICORE system:the UNICORE sends a request for a WS-Agreement based on the job require-ments submitted by a user, theMetaScheduler negotiates the agree-ments through WS-Negotiation anddelivers the resulting WS-Agreementsback to the UNICORE system.

Finally, another objective of VIOLA isto connect to similar projects on aEuropean and international level; theseinclude the Canadian CANARIE, theCzech CESNET, the Polish PIONIER,the British UKERNA and the DutchSURFnet. Efforts to establish theseconnections and create joint projects willbe made starting in 2005.

Links:VIOLA: http://www.viola-testbed.de

WS-Agreement/WS-Negotiation: http://forge.gridforum.org/projects/graap-wg

UNICORE: http://www.unicore.org

Please contact:Helmut Grund, Wolfgang Ziegler,Institute for Scientific Computing and Algorithms SCAI, Fraunhofer ICT Group,GermanyTel: +49 2241 14 2298/2258 E-mail: [email protected], [email protected]

Figure 2: MetaScheduler Integration in UNICORE.

Unicore Client

Unicore Gateway

(Primary) NJSNJSNJS

TSI TSI TSI

Local Scheduler Local Scheduler Local Scheduler

VSITE A VSITE B

MetaScheduler

NegotiationInitiator

Unicore Gateway

WS-Negotiation/Agreement

MPI metascheduling request+ Scheduler URIs

NegotiationResponder



AJO

WS-Agreements



In spite of the success of Grid computingin providing solutions for a number oflarge-scale science and engineering prob-lems, the problem of GRM remains anopen challenge. GRM problems includethe management of user requests, theirmatching with suitable resources throughservice discovery, and the scheduling oftasks on the matched resources. A highlevel of complexity arises from thevarious issues GRM must address:resource heterogeneity, site autonomyand security mechanisms, the need tonegotiate between resource users andresource providers, the handling of theaddition of new policies, scalability, faulttolerance and QoS, just to cite the mostrelevant.

Software agents are emerging as a solu-tion for providing flexible, autonomousand intelligent software components.Multiagent Systems (MAS) arecollections of software agentsthat interact throughcooperation, coordination ornegotiation in order to reachindividual and common goals.There are five main dimensionsthat software computing ispresently trying to address andthat are shared by the Agentperspective:• ubiquity • interconnection• intelligence (and learning)• delegation (autonomic

computing)• human orientation.

Agents and MAS meet theserequirements, and many

computer scientists consider Agent-oriented programming to be the nextparadigm in software computing. Gridcomputing pioneers, as the developers ofthe ‘de facto’ Grid middleware GlobusToolkit, agree with agent experts on theadvantages that the convergence of thetwo approaches may bring (see theirrecent paper: I. Foster, N. R. Jenningsand C. Kesselman, (2004). Brain meetsbrawn: Why Grid and agents need eachother. Proc. 3rd Int. Conf. onAutonomous Agents and Multi-AgentSystems, New York, USA)

The GridCat group at the PolytechnicUniversity of Catalonia (UPC,http://www.gridcat.org) has beenworking for the last two years on a Gridcomputing domain. This will provideGrid services to the research communityformed around the i2CAT project and

Foundation (http://www.i2cat.net),addressing several aspects of Gridcomputing. For example, Globus Gridmiddleware has been tested and extended,and a Grid testbed has been created onUPC resources. We are also involved invarious research activities, includingdeployment of applications on grids, theconstruction of a generic Grid Frameworkand Portal, as well as the current work wepresent here to provide an agent-basedarchitecture for GRM. Aside from these,GridCat is involved in a joint project withSun Microsystems Spain, to develop anation-wide Grid Infrastructure forcollaborative science projects. The busi-ness and industrial sectors are alsoactively cooperating in the design, devel-opment and deployment of grids for themedia sector in the Barcelona area.

We are currently looking at the intermin-gling of Grid and Agent tech-nologies. We propose an envi-ronment based on an agentmarketplace, in which agentsrepresenting Grid users candetect and meet agents repre-senting Grid services, nego-tiate with them for the condi-tions of the service, andexecute the tasks on the Gridnode corresponding to theservice (see Figure 1). Agentsallow flexible negotiation,exchanging messages onbehalf of the entities theyrepresent. In our system,agents can use an extra infor-mation source for negotiations,consisting of a table containingthe possible grid configura-

by Isaac Chao, Ramon Sangüesa and Oscar Oardaiz

Grids and agent communities offer two approaches to open distributed systems.Grids focus on robust infrastructure and services, and agents on the operationof autonomous intelligent problem solvers. The interests of grids and agentsconverge when efficient management of the services provided by grids comesto the fore. Cooperation, coordination and negotiation are issues that Grid usersand Grid resource providers need to address in order to successfully interoperate.Here we present work developed by the GridCat group in order to provide anagent-based Grid Resource Management (GRM) system compliant with adoptedstandards both for both Grid and Agent computing.

Grid Resource Management using Software Agents

Figure 1: The Overall Architecture.



tions ranked by estimated utility. Foreach possible configuration, this utilitytable stores the expected time of taskcompletion, given values for parametersused to model the state of the resources(eg available resources, system load,network traffic) and for executionparameters of the task (eg granularityand time between each submission of asingle job). The Broker Agent (BA)representing the Grid user can check thisutility table to decide which of the bidsproposed by Seller Agents (SAs) is thebest, and also decide how to prepare thetask for the effective execution in orderto minimize the time to completion.

The negotiation between agents forusage of resources is based on the FIPA-Contract-Net protocol, a commonprotocol used to dynamically create rela-tionships in response to currentprocessing requirements embodied in acontract. The contract is an agreementbetween a manager (the Broker) and acontractor (the Seller Agent acting onbehalf of a Grid resource), resulting fromthe contractor successfully bidding forthe contract.

The agents are implemented using JADEFramework supporting FIPA standardfor agent architectures, and they behavein a coordinated manner by means of the

sequential execution of their behaviours(see Figure 2). We use Globus Toolkit 3as Grid middleware for the interactionbetween agents and the Grid services andfor the effective scheduling of jobs intothe available resources.

Future work will cover the inclusion ofeconomic models in the negotiations, aswell as an additional layer of evolu-tionary learning embodied in the agents

in order to efficiently build the utilitytable, thereby allowing the architecture tobe ported to real-world operative grids.

Link:http://www.gridcat.org

Please Contact:Isaac Chao, Ramon Sanguesa, Oscar Oardaiz, Polytechnic University of Catalonia, Barcelona, SpainTel: + 34 93 401 71 11E-mail: [email protected], [email protected], [email protected]

Figure 2:

The Agents

operation

Sequence

Diagram.

The Computational Grid is a piece ofinfrastructure formed from a variety ofcomputational resources interconnectedby wide area network (WAN) links. Thecurrent Internet cannot adequatelysupport Grid services, however, becauseit is only able to offer ‘best effort’service with no resource reservation.This type of service is adequate for tradi-

tional applications like FTP, Telnet andemail but is intolerable for emergingreal-time, multimedia and Grid applica-tions (eg Internet telephony, video-conferencing, video on-demand etc),which require high throughput (band-width), low latency (delay and jitter) andlow packet-loss rate (reliability). Thiscalls for the need to provision Quality-

of-Service (QoS). Provisioning of QoSin a network requires its definition anddeployment using various mechanisms.Network QoS can be defined as ‘a set ofservice requirements to be met by anetwork while transporting a flow’. Therequirements can be qualitative or quan-titative. Qualitatively, QoS is definedthrough user perceptions and require-

by Eric Ayienga , Bernard Manderick, Okello-Odongo and Ann Nowe

Research carried out in cooperation between University of Nairobi and the VrijeUniversiteit Brussel proposes to use multi-agent systems (MAS) for provision ofQoS at the network level in a Grid environment. In the proposed architecture, self-interested agents will interact using electronic market-based techniques with theaim of establishing and maintaining a certain level of QoS in the Grid network.

Multi-Agent Systems for Efficient Quality of Service Routing in Grids



ments. Quantitatively QoS is definedthrough the specification of require-ments in terms of constraints on variousquantifiable metrics such as bandwidth,delay, delay jitter, reliability and cost. Itsdeployment is done through admissioncontrol, scheduling, routing, flow-control and policing. Attempts to satisfythe service requirements are being madeby IETF, IEEE etc, through BandwidthOverprovisioning, QoS Architecturesand Traffic Engineering.

Since ‘best-effort’ will continue to bedominant however, all QoS mechanismsare layered on top of the existing Internetrather than replacing it with a new infras-tructure. The reason for this dominanceis that the infrastructure already exists,and the routing protocols and algorithmsare reliable and stable for the applica-tions it was designed for.

Two of the key issues in supporting theseQoS campaigns in communicationnetworks are QoS specification androuting. QoS specification specifies therequirements needed for QoS and quan-tifies them as accurately as possible.QoS routing on the other hand isconcerned with routing traffic such thatthe QoS requirements of the carriedtraffic are met. In trying to find the‘shortest path’ through the network, thecurrent Internet routing protocols(OSPF, RIP, BGP) have two limitations.First, they use single objective optimiza-tion algorithms which consider a singlearbitrary metric such as hop count oradministrative weight during the pathselection process. This may lead tocongestion on this path while alternatepaths with acceptable but non-optimalconditions are relatively free. Second,‘best-effort’ algorithms tend to shifttraffic from one path to a ‘better’ pathwhenever such a path is found. QoS-based routing should be designed toavoid these problems and in additionshould take into account the applications'requirement and the availability ofnetwork resources. However, QoSrouting poses several challenges thatmust be addressed to enable the supportof advanced services in the Internet, bothat intra and inter-domain levels.

Contemporary trends are movingtowards the provision of QoS through

intelligent resource allocation whilerouting traffic using machine-learningtechniques. This research will approachthis challenge through the use of multi-agent systems (MAS). The motivationthe use of MAS is that grids are opensystems. In these systems, control isdistributed, characteristics are not knownin advance, and are dynamic and hetero-geneous. MAS can address this problemby computing solutions locally based onlimited information from isolated partsof the system, and use this information ina social way. Such locality enablesagents to respond rapidly to changes inthe network state, while their socialnature can potentially enable theiractions to be coordinated to achievesome wider and socially desirable effect.MAS have been known to solve prob-

lems that have the property of inherentdistribution (physically and geographi-cally). They also solve problemsrequiring the interconnection and inter-operation of multiple autonomous, self-interested legacy systems.

In the proposed architecture, agents willinteract in the Grid network infrastruc-ture with the aim of establishing andmaintaining a certain level of QoS. Inallocating resources to implement QoSin Grids, the MAS properties to beexploited include the following:• Agency: agents act as representatives

for other entities with the expresspurpose of performing specific actsthat are seen to be beneficial to therepresented entity.

• Autonomy: agents will makeautonomous decisions based on theirintelligence and social ability.

• Interaction: agents will interact withusers, other agents and the environ-ment through coordination (this can becooperation or competition), commu-nication, and negotiation.

The proposed MAS-based solution willbe decentralized and the predominantMAS structure will be the market; in thiscase the Grid environment in which therewill be buyers and sellers of resources.Link agents will be the producers of theresources in this economy. They willwish to maximize the income they get byselling network resources to path agents.Path agents will be the buyers of linkresources and sellers of path resources.Negotiations between buyer and selleragents will be based on the marketeconomy and will be modelled usingevolutionary game theory. The MASarchitecture is as shown in the Figure.

When a request for a service is made, theInterface Agent will map the servicerequest to the QoS parameters and come

up with a quantitative value for theservice. This value will be used to deriveanother quantitative value for theresources. which the Path Agent will useto competitively negotiate for a link withthe Link Agents. Upon agreement, thePath Agent will use this link to move tothe next node and repeat a similarprocess. In this way a route with the rightQoS parameters for a service request willbe formed through the network.

Initially, the system will be modelledthrough simulation. A simulator repre-senting the current Internet will be built,and the MAS will be built on top of thissimulator as an additional layer ofcontrol. Challenges include finding suit-able paradigms for agent autonomy,multi-agent communication techniques,agent communication platform andlanguage, and the many layers involvedthat will lead to a high communicationand protocol overhead.

Please contact:Bernard Manderick, Vrije Universiteit BrusselE-mail: [email protected]

The MAS

Architecture.


42

Grid computing technology enables thecollaborative usage of computationalresources across different organizations.Nowadays, the Globus Toolkit repre-sents the industrial standard for Gridcomputing, and is involved in a largenumber of research projects. Eventhough Globus offers all the necessarytools to perform Grid-based executions,its steep learning curve means an intensetraining process is required before itsbenefits can be fully exploited in eachparticular area of interest.

In the framework of the project GRID-IT(TIC2003-01318), funded by theSpanish Ministry of Science andTechnology, a middleware has beendeveloped by the Networking and Highperformance Computing ResearchGroup (GRyCAP) of the ValenciaUniversity of Technology in order toprovide easy and transparent access toGrid computing facilities.

Purpose and Description of the MiddlewareThis middleware offers an object-oriented, high-level application

programming interface which allows theprocess of remote task execution in aGrid deployment to be simplified. It isintended to give support to parametersweep applications, which involve theexecution of multiple instances of a task.These sorts of executions are typicallyresource-starved and thus Gridcomputing offers significant benefits byenlarging the computational capabilitiesof a single organization with newcomputational resources from abroad.

As represented in Figure 1, the middle-ware was developed on top of the JavaCommodity Grid Kit, allowing the userto completely avoid direct interactionwith Globus services such as GRAM(for job execution), GridFTP (for high-performance file transfer), or MDS (forresource characteristics discovery).Instead, the user is provided with a high-level API that exposes a very natural andconvenient point of entry to the Gridservices.

It is important to point out that, with thedirect interaction with Globus, Grid usersmust combine all the services provided by

the toolkit in order to achieve theirpurpose, which generally representsremote task execution. This implies thatthe users must concentrate on how toperform the execution, rather than onwhat to execute, which is in factwhat theyare interested in. Since it is the purpose ofthe Grid to offer advantages to users, theeasier it is to use this technology, thesooner the benefits are achieved.

Figure 2 is a diagram showing some ofthe most important classes of the middle-ware. The user directly interacts withinstances of these intuitive classes. Forexample, a GridTask represents anabstraction for a task that must beexecuted in the Grid, a GridResourcerepresents an abstraction for a computa-tional resource in the Grid infrastructureand the BasicScheduler providesscheduling capabilities for the allocationof GridTasks to GridResources.

With this middleware, the user need onlydescribe the application and thesupporting classes involve the entireinfrastructure to achieve the remote taskexecution.

by José Miguel Alonso, Vicente Hernández, Germán Moltó

The Globus Toolkit is considered the de facto standard for Grid computing.However, its steep learning curve has discouraged its widespread adoption. Inorder to overcome this disadvantage an object-oriented software has beendeveloped, which provides an intuitive interface that completely hides the directinteraction with Globus.

Enabling High-Level Access to Grid Computing Services

Figure 1: Middleware location with respect

to the applications and Globus Toolkit.

Figure 2:

Simplified

class

diagram

of the

middleware.

APPLICATIONS

Java CoG Kit

GRAM MDS GridFTP

Globus Toolkit

Middleware Designed




For each task, the middleware allows thedependent input file set to be describedand appropriately staged into theresource before execution. The outputfiles can also be specified, and they willbe transferred back to the local machineafter execution. The task may alsospecify an a priori quality of service,requesting a minimum amount of avail-able RAM or a minimum number of

processors in the remote resource, andrefusing to be executed if these are notfulfilled.

The middleware also provides fault-tolerant dynamic scheduling capabilitiesfor the allocation of tasks to resources.Fault tolerance is achieved by the appli-cation-dependent checkpointing support,periodically retrieving the checkpointfiles to the local machine, so that anexecution can be resumed in anotherresource in the case of failure.

ApplicationsThis middleware is currently being usedby the GRyCAP in a number of compu-tational fields. First of all, in the area ofbiomedical applications, the study of thecardiac electrical activity, especiallyunder pathological conditions, requiresthe execution of several parametricsimulations in order to analyse, forexample, the influence of certain anti-arrhythmic drugs.

On the other hand, the structural analysisof buildings in civil engineering designcan require the simulation of a largenumber of different structural alterna-tives under diverse load conditions, inorder to find that which best accom-plishes with all the economic limitations,design aspects and safety requirements.

In conclusion, the development of high-level interfaces to Grid computing facili-ties greatly enhances the usability of andinterest in this technology, allowingwider adoption of the Grid by non-experts.

Link:http://www.grycap.upv.es

Please contact:José Miguel Alonso, Vicente Hernández, Germán Moltó, Universidad Politécnica de Valencia/SpaRCIM, SpainTel: +34 963 87 73 56E-mail: {jmalonso,vhernand,gmolto}@dsic.upv.es

Figure 3: Principal steps performed to

achieve remote task execution.

File Stage In

Parallel Execution

Checkpoint (Optional)

File Stage Outhp w orkst ati on xw 4000

Local MachineRemote Machine

Creating applications by combiningindependently-managed (oftendistributed) services forms a key themein Grid computing. Such services may behosted by different institutions, andutilise a variety of compute and dataresources. Workflow tools play animportant role in helping compose theseservices, and considerable effort hasbeen put into creating workflow toolsthat combine ease-of-use with supportfor the widest range of Grid middleware.Most Grid-centric workflow toolsassume the existence of a static set ofservices, available at the time oflaunching the tool. To exploit fully thedynamic aspects of Grid infrastructure, itis necessary to allow for dynamicservices to be deployed and discovered.

In automated discovery techniques,services have a rich semantic descriptionand a matchmaking algorithm attempts tomatch the service request with serviceadvertisements. The match is then classi-fied as an exact match, or several flavoursof ‘relaxed’ match. It is important to notethat often ‘automated’ discovery alsorequires having a human in the loop — asit is seldom possible to specify user needsin an exact way (often users are also notfully aware of their requirements). It isintended that the identification of ‘suffi-ciently similar’ services to a query is oftenlikely to lead to a discovery of more suit-able services. The ‘Grid EnabledNumerical and Symbolic Services’(GENSS) project is an EPSRC-fundedproject between the University of Bathand the Welsh eScience Centre (WeSC)

at Cardiff University. The project buildson the development of Mathematical WebServices in the European Commissionfunded MONET (http://monet.nag.co.uk/)project. The GENSS project addresses thecreation of matchmaking techniques toallow users to discover Mathematicalservices described in terms of anOpenMath-based ontology. Mechanismsto allow such services to be delivered aspart of a workflow for mathematicalproblem analysis is also being undertakenin the project. Such service discovery, bynecessity, needs to be problem-specific.By focusing on mathematical services, weintend to address the requirements of amuch wider range of users in science andengineering —than perhaps by focusingon a particular application domain (suchas BioSciences, for instance).

by Omer Rana, Luc Moreau and Julian Padget

Three Projects (PASOA, GENSS, SERENITI) extend the capability of workflowsystems in Grid environments — generally focusing on the capacity to enablemore reliable, trustworthy and efficient use of Grid resources. Such issues arealso significant to encourage greater use of Grid infrastructure bycommercial/business users.

Trusted Grid Workflows: Assigning‘Reputation’ to Grid Service Providers



Once suitable services have been discov-ered, it is often difficult for the user toascertain if the service selected willperform as expected. This introduces theissues of trust and reputation in servicesmade available over Grid infrastructure.Trust issues may be considered from theservice provider's or the serviceconsumer's perspective. For theprovider, trust has always been associ-ated with security -- focusing on authen-ticating and authorising users accessingthe provider's resources. For theconsumer, trust is considered as apredictability issue addressing concernssuch as which service provider is trulyreliable in delivering a service, and howmuch credence should be given to thatprovider to deliver the service as adver-tised. The concepts of trust and reputa-tion are complex and many-facetedissues — and associated with themessuch as risk, competence, beliefs/percep-tions, utility/benefit and expertise. Twoprojects at WeSC are attempting toaddress these concerns: • PASOA (Provenance Aware Service

Oriented Architecture), led by the

University of Southampton (EPSRC-funded)

• SERENITI (SERvice-ENabledInteractive Trust Informatics), led byWeSC.

The PASOA project is investigating howprovenance data may be associated witha workflow process (process prove-nance) and with each service (serviceprovenance). The problems of deter-mining the origin of a result, especiallywhen it involves processing through anumber of stages, or deciding whenresults of analysis are no longer validbecome important concerns in open Gridenvironments and in establishing thequality and reproducibility of scientificapplications. The PASOA project there-fore is aiming to define execution andservice provenance in relation to work-flow enactment, to enable better use ofthis data for reasoning about the‘processes’ involved in a scientificcollaboration over the Grid.

The SERENITI project on the other handis exploring how the concepts of trust

and reputation are related to each other,and especially how a reputation index’may be assigned to a group of services,based on the reputation of each servicewithin the group. The key theme in theproject is the observation that in a work-flow session there is often a set of ‘crit-ical’ services which must performsuccessfully. These critical services arealso only available from a restricted setof providers (such as a particularnumeric solver that can only execute ona restricted set of platforms). Allocatinga reputation-index to such criticalservices, and also determining how thisvalue changes over time is a key factor inassessing risk for the entire workflow.

Links: GENSS: http://genss.cs.bath.ac.uk/ PASOA: http://www.pasoa.org/

Please contact:Omer F. Rana, Cardiff University, Wales, UKTel: +44 29 2087 5542E-mail: [email protected]

Julian Padget, University of Bath,E-mail: [email protected]

Luc Moreau, University of Southampton, UKE-mail: [email protected]

There are many efforts worldwide toprovide new Grid middleware conceptsfor constructing large production grids.As a result, the Grid community is atthestage of producing third-generationGrid systems represented by the OGSA(Open Grid Services Architecture) andWSRF (Web Services ResourceFramework) standards. On the otherhand relatively little attention has beenpaid to how end-users can survive in therapidly changing world of Grid genera-

tions. The primary goal of the researchteams of University of Westminster andSZTAKI within the UK e-ScienceOGSA Testbed Project was to constructa high-level Grid application environ-ment where end-users can: • deploy and use any legacy code

program as a Grid service whencreating Grid applications

• easily and conveniently createcomplex Grid applications byaccessing a Grid portal.

GEMLCA is a Grid architecture devel-oped by the research team at the Centrefor Parallel Computing, University ofWestminster, and aims to make legacycode programs available as Grid serviceswith user-friendly interfaces and withoutre-engineering the original code. UsingGEMLCA, legacy code written in anylanguage (Fortran, C, Java etc) can easilybe deployed as an OGSA Grid servicewithout any user effort. GEMLCA hasalso been integrated with the P-GRADE

by Péter Kacsuk, Tamás Kiss, Ariel Goyeneche, Thierry Delaitre, ZoltánFarkas and Tamás Boczkó

One of the biggest obstacles in the widespread industrial take-up of Gridtechnology is the existence of a large number of legacy code programs that arenot accessible as Grid services. We introduce a new approach – Grid ExecutionManagement for Legacy Code Architecture (GEMLCA) – to deploy legacy codesas Grid services without modifying the original code. By integrating GEMLCAwith a workflow execution-oriented Grid portal, the P-GRADE portal, Grid servicesbased on such legacy code can be applied in complex business processes.

A High-Level Grid Application Environment to Grid-Enable Legacy Code



Grid portal and workflow solutions(developed by SZTAKI), allowing theend-user to construct and executecomplex Grid workflows, using a user-friendly Web interface, on which legacycomponents of the workflow show up asGrid services.

GEMLCA is a general solution for thedeployment of existing legacy codeapplications as Grid services withoutcode re-engineering. The only accesspoint for a client to the GEMLCA archi-tecture is the front-end layer composedof a number of Grid services. Theseservices offer interfaces in order todeploy, query, submit, check the statusof, and get the results back from legacycomputational jobs. As with any otherGrid Service, the front-end layer isdescribed in Web Services DescriptionLanguage (WSDL) and can be used byany Grid service client to bind and makeuse of its functionality through SimpleObject Access Protocol (SOAP).GEMLCA clients can be easily createdusing either the provided universalGEMLCA Stubs for Java or the WSDLfile. The conceptual architecture ofGEMLCA is shown in Figure 1.

Our goal was to enable end-users toapply legacy code as OGSA Gridservices in workflows and to provide auser-friendly interface through which todo this. In order to achieve our goalGEMLCA has been integrated with theP-GRADE Grid portal solution. Theworkflow editor of P-GRADE portalenables the connection of componentjobs into complex workflows through aneasy-to-use graphical environment. TheWorkflow Manager of P-GRADE portaltakes care of executing such workflowson various Grid systems in a user-trans-parent way. The integration ofGEMLCA and the P-GRADE portalallows the user to construct and executeworkflows from legacy codes deployedas OGSA Grid services.

In order to demonstrate the capabilitiesof the solution, a complex workflowsimulating and analysing urban cartraffic was created and executed on fivedifferent sites of the UK OGSA Testbed.Simulators run in parallel at SZTAKI,CCLRC Daresbury Laboratory, andWestminster, Portsmouth and ReadingUniversities. The simulators were fed bya road-network generator and the results

were analysed by a program also runningat Westminster. The applications werelegacy C codes transferred into Gridservices using GEMLCA. The workflowand visualization of execution is shownin Figure 2.

Future work will include the support ofother service-oriented Grid architectureslike WSRF and pure Web services. It isalso envisaged that plug-ins for applica-tion-specific visualizers to the P-GRADE portal will be developed.

Links:GEMLCA: http://www.cpc.wmin.ac.uk/ ogsitestbed/GEMLCA/

P-GRADE: portalhttp://www.lpds.sztaki.hu/pgportal/

Please contact:Péter Kacsuk, SZTAKI, HungaryTel: +36 1 279 6064E-mail: [email protected]

Tamás Kiss, Cavendish School of Computer Science, UKTel: + 44 20 7911 5000, ext 3542E-mail: [email protected]

Figure 1:

GEMLCA conceptual

architecture.

Figure 2: Complex workflow and execution graph for analysing road traffic.



Early Grid middleware and tools empha-sized low-level resource managementaspects. Today, Grid computing hasadopted the notion of ‘services’ as abasic building block for large-scalescientific computations. The focus hasmoved from CPU sharing to definingcomputational resources and function-ality as composable services. Thanks tostandards such as Web services, anincreasingly large number of basic Gridservices are being published. Theyencapsulate both algorithms and datasources, hiding the complexities andpeculiarities of the underlying plat-forms. With services such as basicbuilding blocks, it becomes possibleto rapidly build Grid applications bycomposing these services at a highlevel of abstraction.

At the Information andCommunication Systems ResearchGroup at ETH Zurich we have devel-oped JOpera, a system for rapidservice composition based on avisual composition language. WithJOpera, the data transfers betweenservices (data flow), their order ofinvocation and the necessary failurehandling behaviour (control flow)can be specified with a simple,graph-based, visual syntax. Asopposed to an XML-based approach(eg BPML, BPEL4WS, XL), a visuallanguage can greatly en-hance theunderstandability of Grid computa-tions, as they can be specified byliterally drawing them. Furthermore,at run-time, they can also bedebugged and monitored using thesame visual syntax. In addition tonesting, recursion and reflection, theJOpera Visual CompositionLanguage offers several iterationconstructs specifically targeting Grid

computing. For example, large computa-tional jobs are typically partitioned intosmaller units that can then be executed ina pipelined, parallel or sequentialfashion. JOpera includes simple primi-tives for these operations and allows theprogrammer to precisely control whathappens in the case of failures. JOperacan also suggest the most optimal parti-tioning strategy based on the availableresources.

JOpera offers an open, flexible platformfor Grid service composition. It main-tains a library of re-usable components,which can be dragged, dropped andconnected into a data and control-flowdiagram. JOpera makes very fewassumptions on the properties of thesecomponents, so that the user can freelychoose to compose the most appropriatekind of services in terms of per-formance, reliability, security andconvenience. Such components may

represent the invocation of basic,remote Grid services, but also, forexample, job submissions to externalresource management and schedulingsystems or the execution of localapplications under a variety of oper-ating systems. Additionally,composite Grid services can be re-used in two different ways. On theone hand, Grid computations can bedecomposed hierarchically intodifferent modules, which can beinvoked and re-used independently.On the other hand, composite Gridservices can define re-usable patternsof interaction between services,which can then be customized andtailored to specific applications. Todo so, service interfaces are bound toactual Grid sites providing compat-ible services at deployment or invo-cation time.

For efficient execution, the JOperaVisual Composition Language iscompiled into Java code. To ensurethe required level of scalability, relia-bility and flexibility, the resultingcode is dynamically plugged into aruntime container: the JOpera kernel.The flexible architecture of theJOpera kernel can be deployed indifferent configurations, eg stand-alone or embedded into other

The recent shift to service-based Grids enables the use of service compositiontools for rapidly building and efficiently running distributed computations. At ETHZurich, we have developed JOpera, a visual composition language and a set ofintegrated software development tools for composing Grid services. The systemcan be freely downloaded and has been successfully tested with data-intensivebioinformatics applications as well as large-scale, Monte-Carlo networksimulations.

JOpera: Visual Composition of Grid Services by Cesare Pautasso

Architecture of the JOpera system.

Transform Input

Transform Output

ComputeDistances

CalcStatistics

ComputeRoutes

high

GridService Input

GridService Output

DownloadInputData

Transform

UploadResult

CalcStatisticsComputeDistances DownloadInputData UploadResult

Visual Composition Language and Tools

Java Code Monitoring API

Compiler Debugger

JOpera Execution Kernel

Grid Service Library

RemoteGrid Service

(GLOBUS)

Local Clusterof Computers

(Condor)

DatabaseServer(JDBC)



systems, such as development environ-ments or application servers. An openAPI is a key feature for providing accessto composite Grid services in a variety ofways. In the simplest case, once Gridservices are composed, the resultingcomputation can also be published as aGrid service. To do so, the JOpera kernelmanages the execution of the compositeservice behind a Grid portal. Similarly,JOpera can be embedded into end-user

applications, thereby supporting thedevelopment of application-specificfeatures built by composing existingGrid services. Additionally, we havedeveloped generic, visual monitoringtools for keeping track of the progress ofa Grid computation. In order to handlelarge workloads, the JOpera kernel canscale across a cluster of computers. Weare currently investigating how to applyautonomic techniques in order to extend

the system with a controller in charge ofdetermining the optimal configuration ofthe system based on the current work-load.

Link:The latest version of JOpera can be freely downloaded from http://www.iks.ethz.ch/jopera

Please contact:Cesare Pautasso, ETH Zurich/SARIT,SwizterlandTel: +41 1 632 0879E-mail: [email protected]

Weka is a widely used toolkit formachine learning and data mining origi-nally developed at the University ofWaikato in New Zealand. It is a largecollection of state-of-the-art machinelearning algorithms written in Java.Weka contains tools for classification,regression, clustering, association rules,visualisation, and data pre-processing.Weka is open source software under theGNU GPL. It is easily extensible, whichallows researchers to contribute newlearning algorithms to Weka, keeping itup-to-date with the latest developmentsin the field. As a result, Weka hasbecome very popular with academic andindustrial researchers, and is also widelyused for teaching purposes.

The main focus of Weka is on classifieralgorithms. Simply put, a classifier mapsa set of data instances onto a finite set ofclasses. Each data instance is describedby its attribute values. For example,predicting whether it is going to rainbased on observations of sky, air temper-ature, humidity, and wind can be viewedas a classification task. Each datainstance includes values of the observa-tion attributes, eg (sunny, warm, humid,strong), and the available classes are{rain, dry}. The goal of classifier

learning (or training) process is to derivea classifier from a set of labelled data (iea set of data instances together with theircorrect labels). The idea is that a classi-fier learned on a labelled data set canthen be used to predict class labels forfuture (unlabelled) data instances.

Learning a classifier and using it to labeldata can be time consuming and requiresignificant amounts of memory, espe-cially for large data sets. Unfortunately,parallelising and distributing classifierlearning is a difficult research problemon its own. Nonetheless, there are anumber of simpler functions and usagescenarios in Weka that can still benefitfrom distributing the work on a Grid.The most obvious are labelling, testing,and cross-validation functions.

Labelling involves applying a previouslylearned classifier to an unlabelled dataset to predict instance labels. Testingtakes a labelled data set, temporarilyremoves class labels, applies the classi-fier, and then analyses the quality of theclassification algorithm by comparingthe actual and the predicted labels.Finally, for n-fold cross-validation alabelled data set is partitioned into nfolds, and n training and testing itera-

tions are performed. On each iteration,one fold is used as a test set, and the restof the data is used as a training set. Aclassifier is learned on the training setand then validated on the test data.

In Grid-enabled Weka, which is beingdeveloped in University College Dublin,execution of these tasks can bedistributed across several computers inan ad-hoc Grid. The labelling function isdistributed by partitioning the data set,labelling several partitions in parallel ondifferent available machines, andmerging the results into a single labelleddata set. The testing function isdistributed in a similar way, with teststatistics being computed in parallel onseveral machines for different subsets ofthe test data. Distributing cross-valida-tion is also straightforward: individualiterations for different folds are executedon different machines. The quality of aclassifier often depends on various algo-rithm parameters. The same classifiermay need to be trained with differentparameters to obtain better results. In oursystem, the user can run a training taskon a remote machine. This allows thesame user to train several classifiers inparallel by launching multiple Wekatasks from the user's computer.

by Rinat Khoussainov, Xin Zuo and Nicholas Kushmerick

In the end of the day, Grids are about Grid-enabled applications. While a numberof general-purpose libraries for scientific computing have been adapted for Grids,many more opportunities remain in various specialist areas. We describe herean ongoing work on Grid-enabled Weka, a widely used toolkit for machine learningand data mining.

Grid-enabled Weka: A Toolkit for Machine Learning on the Grid



The two main components of our systemare Weka server and Weka client. Theserver is based on the original Weka.Each machine participating in a WekaGrid runs the server component. TheWeka client is responsible for acceptinga learning task and input data from a userand distributing the work on the Grid.The client implements the necessaryfunctionality for load balancing and faultmonitoring/recovery. It also allows theuser to specify resource constraints for agiven task and takes these into account

when allocating the jobs to servers. Theserver translates client requests into callsto the corresponding Weka functions. Italso provides additional functions likedata set recovery from local storage aftera crash. The same server can be used byseveral different clients, which allowsusers to share resources of the samemachine.

The system uses a custom interface forcommunication between clients andservers utilising native Java object serial-

isation for data exchange. The obviousnext step is to convert it to an OGSA-style service using, for example, aGlobus API toolkit. An importantadvantage of the current implementationis that in a trusted and centrallycontrolled environment (eg a localcomputer lab) it allows for utilisingidling computing resources withminimal set up, configuration, and main-tenance efforts. This is especially conve-nient for machine learning practitionerswho may not be proficient in parallelcomputing or Grid technologies. TheGrid-enabled Weka is currentlyreplacing the original Weka in Elie, amachine learning-based application forinformation extraction developed inUniversity College Dublin.

Links:Weka: http://www.cs.waikato.ac.nz/~ml/weka/

Elie: http://www.cs.ucd.ie/staff/nick/ home/research/download/finn-ecml04.pdf

Please contact:Rinat Khoussainov, Nicholas KushmerickUniversity College Dublin, IrelandE-mail: [email protected], [email protected]

Grid-enabled Weka, usage scenarios.

Examples include issues such as nuclearwaste encapsulation, adsorption ofpollutants on the surfaces of soil parti-cles, and the interaction of mineralsurfaces and fluids. The simulations weperform range in their complexity. Insome cases we need to simulate systemscontaining millions of atoms, and for thiswe represent the forces between atomsby simply parameterised models. Inother cases we need to have more accu-rate representations of these forces, and

we use quantum mechanical methods,but with smaller numbers of atoms.

At its heart, the challenges faced inscaling up these calculations are compu-tational, but with larger simulationscomes an increased need manage data innew ways. In addition to the fact thatlarger simulations generate larger datafiles, we also have the problem that withhigher throughput of calculations comesthe need for intelligent file managementwithin an individual study. The

eMinerals minigrid has been developedas an integrated compute and data gridwith the aim to equip the scientists towork with more complex simulationstudies than before.

The eMinerals minigrid is built around aheterogeneous set of computingresources, quite deliberately so becausedifferent types of molecular simulationshave very different requirements. Thecompute resources include 4 clusters ofPCs, an IBM p-series parallel computer,

by Martin Dove

The eMinerals project is one of the UK Natural Environment Research Council(NERC) e-science testbed projects (formal name 'Environment from the MolecularLevel'). The scientific aim is to use Grid technologies to push forward thecapabilities of molecular-scale simulations for the study of environmentalprocesses.

The eMinerals Minigrid: An Infrastructure to Support Molecular Simulation Scientists


49

a Condor pool containing a range ofmachines but primarily designed toallow test jobs with high memoryrequirements, and a large (930 machine)Condor pool containing the teaching PCsof University College London. The latteris the largest Condor installation in theUK, and is designed to allow high-throughput calculations that do not havehigh memory requirements. Access tothese machines is through Globus, witheach member of the team having a UKdigital certificate to ensure authorisationand authentication. Moreover, we expectthe scientists to submit jobs using theGlobus toolkit commands rather than viadirect logins to the resources. The onlylogins are permitted to allow code devel-opers to compile and test.

One immediate problem with this regimeis that of file transfer. Both Globus andCondor present problems for users in thisregard. Our solution has been to directlyincorporate the Storage Resource Broker(SRB) developed by the San DiegoSupercomputer Centre. The SRBprovides a means to distribute filesacross a number of file systems (called'SRB vaults') but to be seen by the user asa single file system. The physical loca-tion of files is transparent to the user, andis seen only as a file attribute. We haveset up 5 SRB vaults across the eMineralsminigrid. The integrated compute anddata grid that constitutes the eMineralsminigrid is shown in Figure 1, with thedata component shown in more detail inFigure 2.

The integration of the SRB within theeMinerals minigrid facilitates thefollowing workflow. When a user wantsto run a job, the first task is to upload therelevant files to the SRB. The user thenexecutes a three-stage job using Condor-G, which provides a Condor-style wrap-ping for the Globus commands. The firststage is for these files to be downloadedfrom the SRB onto the computeresources being used. The second stageis to run the job. The third stage is to putall output files into the same directorywithin the SRB for easy access to theuser. These three stages have beenwrapped within a single script written byone of the project members, although weare working on a compute portal to makejob management easier for the scientists.

The final stage in the process is longer-term management of the data. TheDaresbury project partners have devel-oped a data portal that has a direct inter-face with the SRB. Data that requirelong-term storage are uploaded to thedata portal from the SRB, and providedwith metadata that enables use to bemade of the data by the scientist orcollaborators at a later date. In order toensure interoperability of data, we aremaking use of the Chemical MarkupLanguage, one of the established XMLstandard languages.

This report represents work in progress.Components of this work, particularlythe eMinerals minigrid as an integratedcompute and data grid, are now beyondproof-of-concept and are in productionmode. Immediate tasks are to provide thetools such as the compute portal to givethe scientists more help in their use of thefull minigrid. The eMinerals teamconsists of scientists, applications codedevelopers and computer scientists/gridspecialists from the Universities ofCambridge, Bath, and Reading,University College London andBirkbeck College, and the CCLRCDaresbury Laboratory.

Link: http://www.eminerals.org

Please contact:Martin Dove, University of Cambridge, UK E-mail: [email protected]

Figure 1: The structure of the eMinerals minigrid, showing both the hardware and

middleware configurations for the integrated compute and data structures and their

institutional configurations.

Figure 2:

The structure of

the data

component of

the eMinerals

minigrid,

showing how

the various

components

map onto the

infrastructure of

the application

server and

database

cluster.




Arteriosclerosis is a widespread diseasethat manifests particularly in developedcountries. Treatment often involvessurgery, such as the placement ofbypasses that lead the blood aroundclogged arteries to restore normal bloodflow. A surgeon plans these interven-tions on the basis of 3D images obtainedfrom MRI or CT scans. Apart fromconsiderations such as accessibility, theattainable improvement in the bloodflow will determine what type of treat-ment is appropriate. Improvements in thesupport for planning these proceduresare expected to improve their successrate. We have developed a prototypegrid-based system for virtual vascularsurgery, which may be used during pre-operative planning or as a valuable toolin the training of novice vascularsurgeons. The prototype uses advanced

distributed simulations and visualiza-tions to support vascular surgeons inmaking pre-operative decisions.

The Virtual Radiology ExplorerThe Virtual Radiology Explorer (VRE),developed at the University ofAmsterdam, is a grid-based problem-solving environment for virtual vascularsurgery. The VRE contains an efficientmesoscopic computational haemody-namics solver for blood-flow simula-tions based on parallel cellular automata.To convert the medical scans intocomputational meshes, raw scanner datais first segmented so that only the arterialstructures of interest remain in the dataset. The segmented data is thenconverted into a computational mesh.The patient’s blood flow is simulatedusing grid resources. The VRE system

uses a desktop virtual reality environ-ment where the patient’s data, obtainedfrom a scanner, is visualized as a 3Dstereoscopic image together with thegraphical interpretation of the simulationresults. A user can then manipulate the3D images of the arteries. This would bethe virtual surgical procedure, eg theaddition of a bypass. Blood flow in thisnew geometry is also computed, and theuser can then assess the effectiveness ofthe proposed treatment, and might tryother alternatives to optimise the proce-dure.

The medical scanners, the visualizationenvironment and the computationalresources required for the flow computa-tions and visualizations are usuallylocated in disparate geographical loca-tions and distinct administrativedomains. The need for transparent accessto these resources, high efficiency andstrict data security triggered the develop-ment of our advanced grid-based simula-tion and interactive visualization forvirtual vascular surgery. The middle-ware to support this was developed in theCrossGrid project.

CrossGridThe CrossGrid project is orientedtowards compute- and data-intensiveapplications characterized by the interac-tion with a person in the loop. TheCrossGrid pan-European distributedtestbed shares resources across sixteenEuropean sites. One key component ofCrossgrid is the Migrating Desktop(MD) grid portal. The MD produces atransparent user work environment,permitting the user to access gridresources from remote computers. Userscan run applications, manage data files,and store personal settings indepen-dently of the location or the terminaltype. We have incorporated our VREsystem into the Grid via the MD gridportal. We achieved secured grid access,

Medical simulations and visualizations typically require computational power notusually available in a hospital. The University of Amsterdam recently demonstratedVirtual Vascular Surgery (a virtual bypass operation), where large-scale simulationand visualization capabilities were offered as services on the Grid.

Virtual Vascular Surgery on the Gridby Peter Sloot and Alfons Hoekstra

Figure 1: Distributed image-based blood-flow simulation and visualization on the Grid.

From top left to bottom right: A patient is scanned in Leiden, The Netherlands; this

results in a raw image stored in a Storage Element on the Grid (eg Poznan, Poland); this

is segmented, filtered and cropped, using a Grid service; a bypass is added; a

computational mesh is generated (on the local machine) and given to the parallel flow

solver, running on Compute Elements in the Grid (Amsterdam and Spain); the resulting

flow fields are displayed on the local machine using visualization services offered by

the Grid Visualization Kernel (Linz, Austria).



node discovery and registration, griddata transfer, application initialization,medical data segmentation, segmenteddata visualization, computational meshcreation, job submission, distributedblood-flow visualization, and bypasscreation. VRE runs on a local machinebut is launched and initialized throughthe MD. The input takes the form ofsegmented or non-segmented medicaldata produced at the Leiden MedicalCentre (LUMC); the CrossGrid testbedprovides access to this data from amedical image repository acting as aGrid Storage Element in Leiden.

A Virtual Bypass Operation on the GridWe have recently demonstrated thefollowing scenario: The abdominal aortaarea of a patient is scanned, and theresulting image is stored in a RadiologyInformation System repository. Later, aphysician (user) logs into the CrossGridPortal using his Grid certificate andprivate key. The user checks if there aresegmented or non-segmented medicaldata ready for analysis in one of thevirtual nodes, and securely transfers afew to his local machine. The user thenstarts the VRE from within the MD,loads the segmented medical data,selects a region of interest, crops theimage, adds a bypass, and creates acomputational mesh. The user selects theBiomedical Application icon within theMD (with parameters and files being

taken from the user's profile), andsubmits the job to the CrossGrid testbed,to the nearest or most adequateComputing Element in the Grid. Theuser may then check job submission orprogress via the MD. After the job hasbeen completed, the calculated veloci-ties, pressure, and shear stress are trans-ferred to the local Storage Element or tothe Grid Visualization Kernel to berendered and reviewed by the user. Wewill give a live demonstration of thisprocess during the upcoming EuropeanGrid Conference in Amsterdam,February 2005.

Overall, the successful deployment ofgrids requires the re-design of algo-

rithms to support loosely coupledcomputer resources. This should bedriven by relevant and challenging appli-cations. With the experience gainedwithin CrossGrid from the virtualvascular surgery application (and threeother applications), European GridTechnology leads the way in new Griddevelopments worldwide.

Links:http://www.science.uva.nl/research/scshttp://www.crossgrid.org/

Please contact:Peter Sloot, University of Amsterdam, The NetherlandsTel: +31 20 525 7462 E-Mail: [email protected]

Figure 2: The CrossGrid Grid

Portal – the ‘migrating desktop’

– with some of its

functionalities, eg grid log-in

and Grid Proxy creation, virtual

node navigation and Grid data

transfer (via the Grid

Commander),

submission of blood

flow simulations to

the CrossGrid

testbed (via the Job

Submission Wizard)

and monitoring of

jobs running in the

testbed (via the Job

Monitoring Dialog).

Over the years, bandwidth has increasedand become cheaper, computers havebecome more powerful, and the Internet

has evolved into an accepted way ofconnecting computers and computingdevices. As resources are rarely utilised

to the fullest extent, companies havebegun to promote the concept of gridcomputing, which enables the integra-

by Mika Pennanen and Markus Ylikerälä

The rapid development and low price of personal computers make them aninteresting choice for conveying ideas through visualisation. Although real-time3D graphics are adequate for many applications, a single computer is insufficientif photorealistic images are required, not only with today’s technology, but evenin the near future. At VTT Information Technology, grid-based distributedcomputing has been utilised to take advantage of the processing power of off-the-shelf office computers that are idle most of the time. As a result, using therich resources of this existing distributed and networked environment,photorealistic images were produced faster and more cheaply.

Photorealistic Visualisation with Grid-Based Technology



tion of computing power from anywherein the world as easily as users now log onto the Internet, or even use electricity.This also opens up the possibility, forexample, of increasing the renderingcapability of low-capacity devices viathe Internet, on demand. Imagine beingable to plug your computer into the walland tapping into as much processingpower as you need. The promise of suchutility computing is very attractive: itmeans obtaining the right amount ofcomputing power at the right time,employing the latest hardware, softwareand services, and paying only for thecomputing power used. The savingscould be substantial.

VTT Information Technology has partic-ipated in several national and interna-tional research projects related to gridtechnology. During these projects weutilised two of our research areas: utilitycomputing used for calculationpurposes, and visualisation and virtualreality environments. The objective wasto use the grid for visualisation purposesto convey ideas to the interest group, andresearch has focused on distributedcalculations for visualisation purposes.The Globus Toolkit has been used, and agrid called Grix has been developedbased on platform-neutral architecture.During our research on distributed calcu-lations for visualisation purposes, weutilised the processing power of off-the-shelf office computers that are idle mostof the time. With the rich resources ofthis existing distributed and networkedenvironment, photorealistic images were

produced more quickly and morecheaply. To obtain photorealisticimages, techniques such as ray-tracingcan be used. Ray-tracing simulates thetravel of rays of light from the observer’sviewing angle to the light sources. Itshould be noted that the simulation ismade contrary to physical laws in orderto reduce the necessary number of calcu-lations. This is based on the fact that theobserver can sense only some of the rays.The program currently in use is POV-Ray, a well-known ray-tracing program.Unfortunately a single computer – notonly current machines but those of thenear future – has insufficient power torender large or complex photorealisticimages. A grid is therefore used in orderto gain more computing power. Anexample of our research on ray-tracingwith our OpenGL application is shownin Figure 1. With this application theuser can move within a room that haschessboard surfaces, a yellow ball and acanvas. The observer takes a snapshotfrom an arbitrary viewpoint, and aphotorealistic image is rendered with agrid and attached to the canvas.

To measure the effect of the grid withvarying numbers of computers, wemeasured the time it took to render aphotorealistic image. The image wasfrom a model of a ski lodge containingtwo light sources and approximately55,000 polygons. The complexity of thisrendering task varied with the imagesize; Figure 2 shows the results of thesetasks. Depending on the image size, theprocessing time with one computer (PIII

processor) varied from nine seconds to47 minutes. With multiple computers,one computer acts as the controller,while the rest perform the actual ray-tracing calculations. As a result, theperformance of two computers is worsethan one alone, since control data merelyincreases overhead. Using the increasedpower of a moderate number ofcomputers does produce significantimprovements, particularly with largeimages. However, for large numbers ofcomputers, the control data overheadincreases such that the results again dete-riorate. Significant benefits can there-fore be obtained even with a smallnumber of computers. Performance isnot improved by continuously increasingthe number of computers.

Our future work will concentrate onextending these solutions and demon-strating them on Grix. Furthermore, it isplanned to extend and implement thebusiness scenario with payable servicesfor device users.

Please contact:Mika Pennanen, VTT Information Technology, FinlandTel: +358 9 456 5623E-mail: [email protected]

Markus Ylikerälä, VTT Information Technology, FinlandTel: +358 9 456 6082E-mail: [email protected]

Figure 1:

Ray-tracing

from an

arbitrary

point of view

in an

OpenGL

application.

Figure 2: Distribution of ray-tracing calculations.



Grid computing is a recent disciplinerelated to the utilization of large-scaledistributed systems for a given purposeby taking advantage of the rich infras-tructure provided by the Internet. Tounderstand the basic idea that motivatesour research, consider a grid ofcomputers as a huge virtual multi-computer ready for processing, storageand communication. Since a grid can bemade up of a set of geographically sepa-rate networks, enormous computerpower is available for solving complexproblems that are limited in CPU andthat require long delays if solved inmodern computer LANs.

Complex problems that can only besolved in non-polynomial time arise inmost fields of research and are becoming

common in many areas of our lives:telecommunications, economy, bio-informatics, industrial environments,andso on. For such a wide spectrum of prob-lems, heuristics come to the rescue, sinceexact techniques are unable to locate anykind of solution. This area is known asNetworking and Emerging Optimization(NEO), and the GISUM group at theUniversity of Málaga (Spain) is workingon just these techniques.

In short, using Grid technology, our aimis to solve optimization problems that areotherwise out of the scope of researchersdealing with parallel algorithms. Fivetypes of applications are usually identi-fied as related to Grid computing:distributed supercomputing, high-throughput computing, on-demand

computing, data-intensive computingand collaborative computing. At thisstage of our research we are primaryinvolved with the two first topics, iedistributed and high-throughputcomputing. However, we plan to enterthe other three domains by developing anopen optimization service for theInternet, solving data-mining problemsand facing software agent applicationsrespectively. Figure 1 is an example ofthe goal of the present work: computingan exact Pareto front.

Multi-criterion optimization in whichseveral (non-dominated) optimum solu-tions must be found is a promisingresearch field. At present, this field lacksalgorithms that could ensure the compu-tation of the exact Pareto front for ageneral problem. We avoid this inconve-nience by using an enumerative-likesearch that computes all the non-domi-nated solutions in a grid. Later,researchers can use these results to findefficient heuristics achieving similar(optimal) results. We are in the first stageof research, and have solved problems atMálaga using a modest grid of around110 processors. As simple as it sounds,finding the optimum Pareto front isextremely hard, even for small problems,and with most algorithms it is not guar-anteed that the optimum front will belocated for arbitrary problems as we dowith our grid exact algorithm.

We are evaluating the performance ofseveral grid-enabled technologies for ourapplications. Concretely, we have testedGlobus, Condor, Legion, and Sun GridEngine. Of these, the first two seem themost suitable for optimization in thegrid. The simplicity and powerfulprocess management of Condor weregreatly appreciated in setting up a gridfrom scratch within a few weeks. Globusis in this sense more complex to use,

by Enrique Alba and Antonio J. Nebro

Research in the ‘Networking and Emerging Optimization Research Line’ at theUniversity of Málaga aims at solving optimization problems through the utiliza-tion of grid–enabled technologies and large computer networks.

Solving Optimization Problems with Grid-Enabled Technologies

Figure 1: A simple example of a multi-objective optimization problem with the

constraints (left), and its Pareto front (right).

Figure 2: Tracking and managing the optimization algorithms in the grid with Condor.



because users must deal with certificates,installation details and production toolsthat require a time-consuming learningcurve. However, Globus actually easesthe next step of our research inconnecting to foreign grids. In addition,we have tested the network usage andexecution times of heuristic algorithmsand tools like MPI on Globus with verysatisfactory results, and this suggests thatwe could achieve fruitful research withGlobus in the future.

Most of this work has been undertakenwithin the on-going TRACER project

(http://tracer.lcc.uma.es), which is devel-oping advanced algorithms in closeconnection with the Internet. Twosuccessful case studies in TRACERrelating to Grid computing can be cited.Firstly, a grid algorithm was developedto compute the exact Pareto front ofseveral hard multi-objective problems byusing Condor in a grid with more than110 computers(http://neo.lcc.uma.es/Software/ESaM).At present, we are extending the grid toother sites in Spain and Europe.Secondly, grid-aware multi-objectiveheuristics, initially run on Globus, were

developed. We are currently testing newgrid algorithms derived from well-known standard techniques like PAES orNSGA II.

We invite everyone to look at our resultsand contact us with any comments orsuggestions for collaboration.

Link: http://neo.lcc.uma.es

Please contact:Enrique Alba, University of Málaga/ SpaRCIM, SpainTel: +34 952 132803E-mail: [email protected]

DB-Annotator was conceived for theResource Description Framework (RDF)annotation of structured informationsources such as relational data or XML-based service descriptions. Only semanti-cally annotated Grid services (GS)provide a means of finding data orcompute-services that are suitable for thetask at hand. Furthermore, they make thebuilding of workflows (coupled serviceswithin the grid) a realistic goal. DB-Annotator will support several levels ofsemantic annotation to enrich Gridservices, ranging from the services them-selves to the data within the Grid.

The project DB-Annotator (see Figure 2)was designed for universal annotation ofdata that can be represented in tabularform, ie data that can be qualified byunique keys.

The main thrust for the developmentcame from the realization that most inter-esting data within biological databasesresides within free-form text fields and istherefore not easily accesible for datamining. Relational databases (RDB) areusually the only choice for data retrievalwithin distributed organizations.Categorial description of the typed data

residing in RDBs can be derived from thetables (entities) themselves, and from thecolumns (entity attributes) and the cellcontents (instances of attributes). Stillmore fine-grained information comesfrom the notorious ‘description’ fields,but putting this implicit information touse for machine-reasoning is unrealisticat this point.

The goal was thus to provide easy navi-gation through existing ontologies anddata-source-independent RDF annotationof RDB data via drag and drop (seeFigure 2). This data categorization can

One of the major challenges in Grid computing is the semantics-driven retrievalof Grid services and distributed data. DB-Annotator is an annotation tool for datain Grid services developed in a collaborative project between the bioinformaticsdepartment and the department for Web applications at the Fraunhofer Institutefor Scientific Computing (SCAI). Fully annotated data in the Grid is particularlyimportant in bioinformatics.

Bioinformatics in the Semantic Gridby Kai Kumpf and Martin Hofmann

Figure 1: Grid Service- Oriented Architecture. The multiple levels

of semantic annotation within a Grid-Service Oriented

Architecture (gSOA). Path 0 corresponds to the semantic

enhancement of the UDDI service registry via RDF. If no additional

semantic glue is needed, direct access to either data- or compute-

services is provided (1a, 2). When retrieving data, the semantic

description can either stem from UDDI or, more finely -grained,

from semantic annotation of structured (most often relational)

data. 1b/c correspond to retrieval of annotated data; both are

equivalent when there exists a central RDF annotation repository

that stores n:m ontology-class -– (data) - relationships. Whereas

1a provides a complete view ofn the data via full-text search or

keys, 1b/c allow querying by content.



take place on three levels within adatabase (see Figure 3), not counting thedatabase itself. A unique data key thatcan be linked to an ontological categorywill therefore appear as‘database:schema:table[:column:row]’,the latter two key parts being optionaldepending on the depth of the annotation.

This concept of assigning ontologyclasses to data was inspired by existingannotation software for the SemanticWeb. Once a data source has receivedample annotation, data mining will bepossible by employing standard methodson the categorized data. With annotatedRDB, access to data services in the Gridwill improve significantly, since queryingby categories is enabled across severalRDBs at once. Further, every result setwill be enhanced with the respective cate-gory information, a data basis that couldbe used to support further automatedannotation of newly inserted data.

Since DB-Annotator produces standardRDF (subject-predicate-object triples

like ‘entity_xyz is-a category_abc’), itsoutput can also serve to enhance UDDI,a proposal made for the introduction ofcontent-searchable Web services and apossible starting point for GS-workflowconstruction. Instead of the databaseunique keys, the RDF would containqualifiers for the available Grid services.Eventually a semantically enhanced GridService-Oriented Architecture (gSOA –SOA being a term borrowed from theWeb services world) would look likeFigure 1. Full categorization of both GSand the exchanged data leaves no moreroom for ambiguities – a preconditionfor data exchange and manipulationwithin the Grid network of virtual orga-nizations (VO).

Given the bulk of biochemical entitiesthat go under various names and interacton different levels, the relevance of fullyannotated data Grid services in bioinfor-matics, especially systems biology, isenormous.. Until now, ambiguities insynonyms and interactions had to bedetected by human intelligence,–

meaning full-fledgedsystems biologyservices won't besuccessfully auto-mated without propercategorization. SinceDB-Annotator drawson a common set ofagreed ontologies (eggene ontology) for theannotation of both dataservices and data setsor subsets, it canprovide the semanticglue on all levelsrequired.

The project was launched in the depart-ment for bioinformatics at FraunhoferSCAI. It is a collaborative projectbetween the bioinformatics departmentand the department for Web applica-tions. Preliminary work commenced latein 2003 with the aim of producing aprototype that would capture the basicfunctionality of the future system. Theprototype, featuring 1:n mapping of adatabase to several ontologies and RDFflatfile storage, was completed alongwith a detailed requirement analysis thissummer. A new structured developmentfrom scratch has now started and isexpected to be finished early in 2005.Among the things that will be includedin the new phase are n:m relationshipsbetween data and ontology classes andstorage in a single, structured relationalrepository. The product is currentlyintended to provide support for semanticannotation of biological and chemicaldatabases. The possibilities for Gridservices annotation within UDDI areexplored within the semantic gridresearch group under the auspices of theGlobal Grid Forum (GGF).

Links:Globus Toolkit: http://www.globus.org

Semantic Web Services: http://swws.semanticweb.org

Global Grid Forum: http://www.gridforum.org

Gene Ontology: http://www.geneontology.org

Please contact: Kai Kumpf, Institute for Algorithms and Scientific Computing (SCAI), Fraunhofer ICT Group, GermanyTel: +49 2241 14 2257E-mail: [email protected]

Figure 2: DB-Annotator was designed to allow RDF-annotation of

structured data. Navigation through a chosen ontology is

facilitated by a DAG view on the classes and their relationships

(left). Any data that can be represented in tabular form and is

accessible via unique keys can be annotated (upper right). Class

attributes have to be filled in on annotation (lower right). In the

case shown here the user has chosen a listing of available Grid

Services that have to be semantically described. The major field

of application for DB-Annotator however, is the annotation of

tabular data from RDBs.

Figure 3: Selection modes in DB-Annotator. Data entities for

annotation can either be the whole table, one column or one

cell. The data key for annotation is constructed depending

on scope of the selection. Only (multiple) annotation of a

single cell corresponds to an extension of the data model.

Selection of a row corresponds to selection of a record from

the extended data model on the basis of table and ontology

keys and as such need not be considered separately.

56

The CoreGRID Network of Excellence(NoE) aims at strengthening andadvancing scientific and technologicalexcellence in the area of Grid and Peer-to-Peer technologies. It is funded by theEuropean Commission through a grantof 8.2 million Euro for a duration of fouryears. To achieve its objective, theNetwork brings together a critical massof well-established researchers (119permanent researchers and 165 PhDstudents) from 42 institutions — severalof which belong to ERCIM — who haveconstructed an ambitious joint programof activities. This joint program of

activity is structured around six comple-mentary research areas that have beenselected on the basis of their strategicimportance, research challenges andexisting European expertise to developnext-generation grid middleware.

Virtual InstitutesTo ensure progressive evolution anddurable integration, the Network is oper-ated as a European Research Laboratory— known as the CoreGRID ResearchLaboratory — having six institutes. Eachof them is dedicated to the particulardomain identified as of strategic impor-

tance to ensure a durable developmentand deployment of GRID infrastruc-tures: • Institute on Knowledge and Data

Management• Institute on Programming Model • Institute on System Architecture • Institute on Grid Information and

Monitoring Services (WP5)• Institute on Resource Management

and Scheduling • Institute on Problem Solving

Environment, Tools and GRIDSystems.

The Network is thus committed tosetting up this Laboratory and making itinternationally recognised and sustain-able. It is funded by a European grantthat is assigned to the CoreGRID NoEfor a duration of four years to cover theintegration costs, while the network part-ners cover the expense required toperform the research associated with thejoint program of activities. Integration isachieved by the joint execution ofresearch projects operated throughresearch groups within the six institutes,the sharing of a common grid testbed forresearch assessment, the access to acommon communication infrastructureto ensure collaboration and dissemina-tion, and a coherent management frame-work to encourage mobility of senior andpost-doctoral researchers and PhDstudents. In addition to these classicalintegration activities, the network has aproactive role to increase the awarenessof trust and security technologies amongthe network participants.

To guarantee that the expertise and theknowledge gained by Network partici-

CoreGRID: European Research Network on Foundations, Software Infrastructures and Applications for Large-Scale, DistributedGrid and Peer-to-Peer Technologiesby Thierry Priol

A Network of Excellence in the area of Grid and Peer-to-Peer technologiescommenced on 1 September 2004, and will allow European researchers to carryout a joint program of activities at an unprecedented level with the aim ofdeveloping next-generation grid middleware.

Figure 1: CoreGRID partners.


U. Cyprus

INRIACNRSERCIM

CETICUCL

CNR-ISTIINFNU. CalabriaU. LecceU. Pisa

CLPP-BAS

Delft U.Vrije U.

EPFLUASF

FhGFZ-JulichHLRSU. MunsterU.DortmundU. PassauZIB

ICS-FORTH

ICRAL-CCLRCU. BelfastU. CambridgeU. CardiffU. ManchesterU. NewcastleU. Wesminster

KTHSICS

SZTAKI

VTT

U. CoimbraUPC

Masaryk U.

U. ChilePSCCYFRONET




There are several Grid research initia-tives, ongoing or planned, at the nationaland European Community level, aimingat the development of a rich set ofadvanced technologies, methodologiesand applications. If the challenges thatlie ahead are to be solved, enhancedcoordination among the authoritiesfunding these initiatives is needed in

order to achieve critical mass, avoidduplication and reduce fragmentation ofefforts.

From 2002 to 2006, EU funding for Gridresearch and deployment more thandoubled, reaching 275 M in FP6. Forthe same period, an estimate of thefunding for Grid research and deploy-

ment by a number of Member andAccession States (UK, France, Italy, TheNetherlands, Germany, Hungary, Spain,Poland, Czech Republic, Sweden) isabout 300-500 M .

The totality of these initiatives couldprovide the EU with the potential to playa leading world role in Grid technologies

by Marco Vanneschi

GridCoord is a Special Support Action (SSA) of the Sixth Framework Programmeof the European Community. The goal is the co-ordination of European initiativesand research activities in Grid Computing in order to strengthen co-operationamong the funding agencies planning future activities, to enhance collaborationbetween the research and user communities, and to develop visionary nationaland EU programmes and roadmaps enabling Europe to play a leadership role inGrid technologies and applications.

GridCoord – Co-ordinating Europe-wide Initiatives in Grid Research

pants will be of mutual benefit to theEuropean Grid community (bothacademic and industrial), CoreGRIDwill organize a set of activities to spreadexcellence outside the Network. Apublic Web site is an essential mecha-nism for the international Grid commu-nity to present the Network as aResearch Laboratory. A set ofCoreGRID publications can be accessedthanks to this Web site (technicalreports, network brochure and newslet-ters). The Network will organize a set ofthematic workshops and conferences onan annual basis. Training activities are

organized to ensure that CoreGRID is aleading educational source in Europe inGrid and Peer-to-Peer technologies. Adedicated activity under the form of anIndustrial Advisory Board will targetmainly industry and commerce to ensurea strong interaction and involvement.

Finally, CoreGRID is committed toparticipating in collaboration activitiesthat can benefit the other EC Gridprojects. In particular, it will take on aleadership role in the organization oftraining activities and will collaborate onEU grid roadmaps, research inventories

and the establishment of a repository ofreference implementations and gridmiddleware.

Link: http://www.coregrid.net

Please contact:Bruno Le Dantec, ERCIMAdministrative & Financial CoordinatorTel: +33 4 92 38 50 13E-mail: [email protected]

Thierry Priol, INRIA, FranceScientific CoordinatorTel: +33 2 99 84 72 10E-mail: [email protected]

Figure 2: CoreGRID vision of the Next

Generation Grid.

Figure 3: CoreGRID participants.



and applications. National and EUcollaborations have been establishedwith other international players (in theUS and Asian-Pacific area) and withinternational standards organisations.However, if Europe wishes to competewith leading global players, it would besensible to attempt to better coordinateits various, fragmented efforts in order toachieve critical mass and the potentialfor a more visible impact at an interna-tional level. This requires: • co-ordination among the funding

authorities in order to maximise theeffective use of the considerableexisting and potential resources

• collaboration among individualresearchers and the creation ofEuropean excellence and competencecentres

• a visionary research agenda.

The first objective of GridCoord is thusto strengthen co-operation among thefunding authorities in order to better co-ordinate the planning of future activitiesin the field of Grid research, an ERAStrategic objective. A second objectiveis to enhance already existing collabora-tions between researchers and users.Finally, it is our intention to contributetowards the development of visionarynational and EU programmes androadmaps, enabling Europe to play aleadership role in Grid technologies andapplications.

Partners The project brings together nationalresearch programme leaders and repre-sentatives from national funding authori-ties with the aim of ensuring nationalcommitment to future joint activities.The following institutions are partners ofGridCoord:• Italy: University of Pisa, Department

of Computer Science, and Universityof Genova, Department ofCommunication, Computer andSystem Science

• France: Institut National de Rechercheen Informatique et Automatique(INRIA), Sophia Antipolis, andUniversity of Nice, Sophia Antipolis

• UK: Engineering and PhysicalSciences Research Council (EPSRC),and Queen's University Belfast,School of Computer Science

• Germany: University of Stuttgart,High Performance Computing Center(HLRS), and Konrad-Zuse-Zentrumfür Informationstechnik (ZIB), Berlin

• The Netherlands: University ofAmsterdam, Faculty of Science,Institute of Informatics

• Hungary: Computer and AutomationResearch Institute, HungarianAcademy of Sciences MTA SZTAKI

• Spain: Technical University ofMadrid, School of Computer Science;

• Poland: Poznan Supercomputing andNetworking Center

• Sweden: Swedish Agency forInnovation Systems (VINNOVA).

Objectives and Activities To achieve the first objective —improved coordination among fundingbodies, policy makers and leaders ofGrid initiatives — the project will: • develop a comprehensive

compendium of national and EUinitiatives including an analysis ofstrengths, weaknesses and gaps at EUlevel

• identify common interests and syner-gies capable of leading to cooperationbetween different groups, collabora-tion among funding authorities andeventually the planning of commoninitiatives to overcome fragmentation,avoid duplication of efforts and buildthe necessary critical mass

• provide a regular forum at whichparticipants from national fundingauthorities and leaders of national Gridinitiatives can meet at six-monthlyintervals to establish synergy betweenthe different projects and programmes.

To achieve the second objective —enhanced collaboration amongresearchers and users of Grid technology— the project will:• organise general and specialized

workshops • promote and organise

dissemination/education/traininginitiatives both for users and forresearch centres

• investigate the status of standard-isation activities in Grid computingand devise a strategy for Europeanactivities in standardisation.

To achieve the third objective — devel-opment of Grid R&D programmes androad maps — the project will:• develop a coherent and innovative

Grid R&D programme vision inEurope taking into account cohesion,and considering present achievementsand the multinational character of GridR&D in Europe. This task will addressthe strategic research issues identifiedby the EU Expert Group on NextGeneration Grids (June 2003) and theUK Gap Analysis report (May 2003),which have been substantially stimu-lated by several ongoing national andEuropean research initiatives and/orcollaborations at international level

• develop Grid R&D road maps andimplementation strategies on the basisof the Grid R&D programme vision.

The multinational character of Gridresearch in Europe should be maintainedand developed according to the intrinsicstrengths of the different nationalprogrammes. This implies the continueddevelopment of both individual nationaland EU programmes for future Gridresearch initiatives at the same time assupporting multilateral exchange of expe-riences and plans to create a morecomplementary and co-ordinatedapproach towards Grid Research inEurope.

GridCoord is coordinated by theDipartimento di Informatica, Universityof Pisa, Italy; Coordinator: MarcoVanneschi; Project Manager: RomanTirler. The starting date was July 2004,duration 18 months, budget 960,000 ,for a total cost of 1155 .

Link: http://www.gridcoord.org

Please contact: Marco Vanneschi, Dept. of Computer Science, Pisa University, ItalyTel: +39 050 2212738 E-mail: [email protected]


59

Recent years have seen an unprece-dented acceleration in the evolution ofthe Internet as the technological vehicleunderpinning the expansion of serviceprovision and inter-/intra- enterpriseintegration in all market sectors. Thisbrings about the prospect of ad hoc inte-gration of systems across organisationalboundaries to support collaborations thatmay last for a single transaction orevolve dynamically over many years.This sets new requirements for scala-bility, responsiveness and adaptabilitythat necessitate the on-demand creationand self-management of dynamicallyevolving virtual organisations (VO)

spanning national and enterpriseborders, where the participating entities(enterprises or individuals) poolresources, information and knowledge inorder to achieve common objectives.The objectives may be short term - eg todeliver a one-off service in response to aspecific customer demand - or long-lasting. In the latter case, the VO's struc-ture, business processes and operationalinfrastructure must adapt as the goals ofthe collaboration, the participating enti-ties, the business context and the tech-nologies employed, change.

Emerging ICT paradigms such asAutonomic computing, Utilitycomputing and Grid computing aremaking the formation and operation ofvirtual organisations easier by providingdynamic management of the distributionof computational processes across avail-able resources. Notwithstanding themajor ICT breakthroughs of the last twodecades, protecting one’s assets whileintegrating services, processes andresources, remains a major ICT chal-lenge. Overcoming such challengesrequires the development of novel tech-nology realising innovative ideas overwidely acceptable interoperable plat-forms. The required scalability, respon-siveness and adaptability for dynamicvirtual organisations, makes the provi-sion of cost effective trust and contractmanagement solutions for VO environ-ments, the most demanding and timelyresearch challenge in this field. Effectivesolutions require interdisciplinaryapproaches, at a global level, integratingtools from law, cognitive and socialscience in addition to telecommunica-tions and computing.

The TrustCOM ProjectIn response to this challenge, theEuropean Commission and a consortiumof end-users, major software vendorsand telecom operators, and academicresearch organisations, are imple-menting the new Integrated ProjectTrustCoM, which is conducting multi-disciplinary research in order to deliver:

by Theo Dimitrakos, Michael Wilson and Santi Ristol

TrustCoM is a new European integrated project that aims to develop a frameworkfor trust, security and contract management in dynamic Virtual Organisations.The framework will enable the secure enactment of collaborative businessprocesses in self-managed, and dynamic value-chains of businesses andgovernments. The framework will leverage and extend the emerging convergenceof open-standards, and in particular Web Services, and open Grid Servicesarchitectures. Validation will take place within industrial strength test-beds incollaborative engineering and electronic business. This article provides anoverview of the TrustCoM project vision, objectives and anticipated results.

TrustCoM - A Trust and Contract ManagementFramework enabling Secure Collaborations in Dynamic Virtual Organisations

TrustCOM Consortium.




Moving on from the distributed platformbased on a distributed infrastructurescenario, Grid.it places special emphasison the high performance requirements ofapplications developed on the Grid. Thismeans that the integration of systems andresources, as well as heterogeneity anddynamic situations management, mustexplicitly handle the case of Grid nodes(which in general are geographicallydistributed or located on private virtualnetworks) as high performance systems,such as parallel machine architectures orclusters. Research on high performance

extends to all the platform levels, fromhigh bandwidth networks to middlewareservices and, in particular, to resourcemanagement as well as tools andprogramming environments.

At the programming tools and environ-ment level the high performance require-ment implies that, when designing scal-able virtual organisations (VOs), aunifying approach should be used. Thisapproach takes into account both aspectsrelated to the distribution of computa-tions and resources, and those related to

parallelism. The programming environ-ment must therefore be characterized bya high degree of portability on differenthardware-software systems (or differenthardware-software combinations) in aheterogeneous and dynamic context.Portability must be guaranteed not onlyfor code but should also ensure that theperformance matches the configurationof the target system at hand. Tool inter-operability and high performance appli-cation reuse are also fundamental to theprogramming environment.

by Domenico Laforenza and Marco Vanneschi

Grid.it is a strategic project in the area of Enabling Technologies for the InformationSociety, coordinated by the Italian National Research Council. The project has astrong interdisciplinary character and is aimed at defining, implementing andapplying innovative solutions for networked computing-enabling platforms,oriented towards scalable virtual organisations and based on the 'Grid Computing'paradigm. The research topics in the project span from high performance photonicnetworks to innovative middleware services, high-performance programmingenvironments and so forth.

Grid.it – Next Generation Grid Platforms and their Applications

• a novel trust and contract managementreference architecture that will enablecollaborative work within on-demandcreated and self-managed dynamicVOs leveraging on the emergingconvergence of Web Services andGrid technologies

• a set of conceptual models explainingthe fundamental concepts, principlesand methods underpinning the abovearchitecture. Effectively these providethe meta-model of any new architec-tural constructs that may result fromTrustCoM research

• a set of profiles, that bring togetherand potentially extend selectedWeb/Grid Services specifications atspecific version levels, along withconventions about how they worktogether to support potential imple-mentations of the TrustCoM frame-work

• a reference implementation of theabove integrating and extendingalready established or emerging inter-operability standards for autonomic

security, trust and contract manage-ment based on Web and Grid servicestechnology

• system and software engineering toolsand methods analysis the VO life-cycle and offering a library of designpatterns and generic software compo-nents implementing selected servicesthat offer the core functionalities of theVO

• testbeds exhibiting instantiations ofthe above architecture and referenceimplementation into two classes ofrealistic application scenarios, namelycollaborative engineering (CE) andprovision of ad-hoc aggregatedservices (ADP)

• selected demonstrators exhibiting thebusiness value and benefits of theTrustCoM framework in the above-mentioned application domains

• studies analysing selected aspects ofthe legal and socio-economic contextthat underpins such VirtualOrganisations.

The TrustCOM ConsortiumThe TrustCoM consortium bringstogether researcher from many majormiddleware vendors and provides abalanced blend of academic and appliedresearchers, end-user organisations, andenterprises looking to utilise results inproducts and services. As such it is well-placed to define, conduct and exploitleading edge research that is relevant tothe needs of European business, govern-ment and society.

Link:http://www.eu-TrustCoM.com

Please contact: Theo Dimitrakos, Scientific Coordinator,CCLRC, UKE-mail: [email protected]

Michael Wilson, Programme Manager,CCLRC, UKE-mail: [email protected]

Santi Risto, Operations Manager, CataluñaAtos Origin,SpainE-mail: [email protected]



At the middleware level, research onresource management includes aspectsof maximum importance such asdiscovery, brokering, scheduling, moni-toring and performance evaluation/prediction.

The study of the high speed networksneeded to support enabling Grid plat-forms for scalable VOs is an internatio-nally recognized 'hot' topic. Within thisresearch activity an important role isplayed by experiments on very highbandwidth optical networks, based onphotonic technology for Grid platformswith high performance sites that extendto the metropolitan area.

In addition to aspects concerningprogramming environments andresource management, the software tech-nology studied in Grid.it includes somefundamental aspects related to middle-ware:• security: secure Grid environments

and cooperation among Grid environ-ments belonging to different organiza-tions

• data intensive services: federateddatabase services, visualization andhierarchical management of data andmeta-data according to advances andhigh-performance techniques

• knowledge discovery services: Gridservices (data mining, search engines,etc.) that provide consistent, efficientand pervasive access to high endcomputational resources

• Grid portals: Grid enabled applicationservices, eg providing the user with thepossibility to submit tasks to remotejobs and collect results via Web inter-faces.

The design and implementation of scien-tific libraries that can be used in a hetero-geneous and dynamic context, such asthe Grid, completes the research onprogramming environments.

ApplicationsGrid.it includes the development ofsome demonstrators selected withinapplication fields that are of maximuminterest, not only for their scientificvalue, but also as testbeds for highperformance Grid platforms:• earth observation• geophysics• astronomy• biology• computational chemistry.

In order to be able to implement andexperiment the ideas and results of theproject, a Grid infrastructure will beimplemented on a national scale, basedon the GARR network and using, insome metropolitan sites, optical fiberinterconnections. This project result willprovide the community with a nationalGrid that can be used in different compu-tational science research sectors, andalso for commerce, industrial and socialservice applications.

OrganisationSix Italian research units (ISTI-CNR ,ISTM-CNR, ICAR-CNR, INFN, CNIT,ASI) participate in the project and alsoinvolve a large number of Italian univer-sity departments.

The principal investigator is MarcoVanneschi, Pisa University and ISTI-CNR, Pisa. The Technical Board is coor-dinated by Domenico Laforenza, ISTI-CNR, Pisa.

The Grid.it project aims at playing animportant role in the training of highlyqualified young people. A relevant partof the project budget is reserved forcontracts for young researchers.

The project has funding of 8.1 M fromthe Italian Ministry for Education andResearch for three years (November2002 – November 2005).

Link: http://www.grid.it

Please contact:Domenico Laforenza, ISTI-CNR, ItalyTel: +39 050 3152992 E-mail: [email protected]

Marco Vanneschi, Pisa University, ItalyTel: +39 050 2212738 E-mail: [email protected]

Research Issues of Grid.it

and System Levels.


Articles in this Section63 Grammar Enables Effective Multimedia Search

Queriesby Fedde van der Lijn, CWI and Menzo Windhouwer,University of Amsterdam, The Netherlands

64 From Textual Use Cases to Behaviour Specificationsby Vladimir Mencl, Charles University/CRCIM, CzechRepublic

65 GHT*: A Peer-to-Peer System for Metric DataMichal Batko and Pavel Zezula, Masaryk University Brno,Czech Republic; Claudio Gennaro, ISTI-CNR, Italy

66 CWI's FACS Cluster - Facilitating the Advancement ofComputational Scienceby Stefan Blom, Sander Bohte and Niels Nes, CWI, TheNetherlands

68 SynDEx: System-Level CAD Software for Optimizing Distributed Real-Time EmbeddedSystemsby Yves Sorel, INRIA, France

69 Large-Scale Simulations of Diffusion in Cell Biologyby Ivo F. Sbalzarini, Anna Mezzacasa, Ari Helenius andPetros Koumoutsakos, ETH Zurich/SARIT, Switzerland

70 Real-Time MIMO Testbed for Next Generation WirelessLANs by Andreas Burg and Helmut Bölcskei




Multimedia is everywhere. Libraries andmuseums digitize their collections andmake them available to interestedparties. Cheap disk space has brought thestorage of large amounts of multimediawithin everyone’s reach. At the sametime however, it has complicated theretrieval of objects.

Not only have databases become larger,they also contain more types of media.While most people are able to find a textcontaining certain keywords, anyonewho has ever used Google Image Searchto look for a specific picture knows thatfinding what you need can be far fromeasy. The majority of search engines canonly interpret textual data. They cannot‘see’ what is depicted in an image or‘hear’ what is on an MP3.

AnnotationOne way of dealing with this problem isto annotate media objects in advance.Annotations describe particular featuresof the stored media objects and can beused to guide semantic search queries.For example, when MP3s are annotatedfor genre and background informationon the performing artist, search querieslike ‘find me all blues songs played byguitarists from Mississippi’ can givemeaningful results.

Annotation can be done manually, butfor large multimedia collections thisquickly becomes impossible. It is there-fore necessary to turn to automatic anno-tation using extraction algorithms. Thesealgorithms can perform easy tasks, suchas finding the length of an MP3, orcomplicated ones, like detecting humanfaces in images. Designing clever extrac-tion algorithms is a necessary conditionfor an automatic annotation system.

Context DependencyYet high-quality extraction algorithmsalone are not enough. Equally importantis a system to coordinate the extractionof annotations. The main problem is thatannotations depend on context, and inpractice this means they depend on eachother. This has consequences for the wayextraction algorithms should be used.For example, feeding charts or logos to aface extractor would be a waste ofsystem resources and time. The annota-tion system makes sure the extractor isonly applied to images with a highchance of containing faces, like photosor drawings.

A lack of context also complicates incre-mental maintenance. Without know-ledge of context and interdependencies,the entire annotation process must bererun every time an extraction algorithmis added or replaced. The annotationsystem finds dependencies and deter-mines exactly which annotations shouldbe updated and which can be re-used.

Context dependency of annotations canalso cause extraction algorithms to giveambiguous answers, especially whendealing with complex features. This canresult in outputs like ‘this image eithercontains a human or a pig’. The annota-

tion system should be able to handlethese ambiguities.

Valid SentencesCWI solves these difficulties bycombining database technology withideas from formal language theory toform the Acoi annotation managementsystem. Its basis is the theory of featuregrammar systems. Such a system notonly describes the annotations them-selves, but also their dependencies andcontexts. Feature grammar can becompared to grammar in naturallanguage. Grammar determines whichword classes can be combined and in

what order to form a valid sentence. Afeature grammar system does the samefor annotations and extraction algo-rithms. It determines what extractionalgorithms must be called to form a validannotation ‘sentence’.

Since the feature grammar system alsostores each annotation’s place in thenetwork of interdependencies, incre-mental maintenance is possible. Whenupdating the database, sentences can bereinterpreted to determine which extrac-tions must be redone. Furthermore, tech-niques from formal language theorycould be modified and used in the anno-tation system. Resolving ambiguities, for

Retrieving files from a multimedia database is like finding a book in a library –without a catalogue of keywords you could search for ages. However, generatingand updating such a catalogue is almost as difficult. CWI introduces so-calledfeature grammar systems to facilitate these tasks.

Grammar Enables Effective Multimedia Search Queriesby Fedde van der Lijn and Menzo Windhouwer

With the advent of digital

cameras, DVDs and MP3s it

has become very easy to

compile large multimedia

libraries. At the same time,

it has become increasingly

difficult to effectively

search collections that

contain a wide range of

media types like text,

images, movies and audio.



example, is a classic problem in thisbranch of computer science.

Case StudiesAcoi has proven its capabilities in avariety of case studies. Together with anumber of basic extraction algorithms itwas used to create an annotation indexfor a collection of Web pages.Furthermore, it was used in combinationwith a presentation generator to unlockthe digitized collection of theRijksmuseum Amsterdam to the public.The generator, also developed at CWI,

uses the annotations to automaticallycompose a semantically structuredmultimedia presentation on a user-defined subject.

CWI’s feature grammar system isunique. Other annotation systems havebeen developed, but all lack the explicitstorage of annotation context. As a result,Acoi is the only system that elegantlyhandles ambiguities and allows for incre-mental maintenance. In the near future,Acoi will be used in the MultimediaNproject in which CWI participates.

Links:http://www.cwi.nl/~acoihttp://www.windhouwer.nl/menzo/professional/index.htmlhttp://www.cwi.nl/ins1

Please contact:Menzo Windhouwer, University of Amsterdam, The NetherlandsTel: +31 20 525 3104 E-mail: [email protected]

Martin Kersten, CWI, The NetherlandsTel: +31 20 592 4066E-mail: [email protected]

The functional requirements for a futuresystem are typically documented withuse cases. In the traditional approach,plain English is used for use cases. Astrong benefit of this approach is that usecases are comprehendible to a wide audi-ence. A textual use case describes thepossible scenarios of the future system,by showing how a particular goal isachieved by the System under design(SuD) and its surrounding actors in asequence of steps. A step of a textual usecase describes an action that is either arequest or information passed betweenan actor and SuD actors, or an actioninternally processed by SuD. Figure 1shows a fragment of a textual use case.

An obvious drawback of writing usecases in plain English is the lack offormal tools to reason upon the require-ment specifications. Also, the subse-quent stages of the software develop-ment process have to start with manuallyprocessing the use cases – eg, identifyingoperations of future objects. Luckily, usecases are typically not written in plain

natural language. The guidelines forwriting use cases significantly restrictthe language by asking for simple anduniform sentence structure; a use casesentence should be in active voice,following the simple pattern 'subject –verb – direct object – prepositionalphrase'. The subject should be the entity(SuD or an actor) active in the task.

Readily available linguistic tools, eg, thestatistical parser developed by MichaelCollins at the University ofPennsylvania, permit obtaining the parsetree for a sentence. While the parsing isnot 100% reliable (it is statisticalparsing), the accuracy improves forsimple sentences – as is the case for usecases. The simple sentence structurepermits us to obtain the principalattributes of the action described by thesentence from its parse tree. With only asimple domain model (names of entitiesand a list of conceptual objects), we canidentify the entity active in a particularstep of a use case. Depending on the roleof the active entity in the use case (SuD

vs. an actor), it is possible to determinewhether the action described by the stepis receiving a request, sending a request,or an internal action of SuD.

We illustrate the approach on the parsetree of step 1 shown in Figure 2. Thesubject of the sentence, 'Seller', a noun(NN), is the name of an actor involved inthe use case. Therefore, the stepdescribes an event (a request) receivedby SuD from the entity Seller.Subsequently, we estimate the eventlabel for representing the request. In theparse tree, we identify the verb (VBZ)'submits' and the direct object 'itemdescription'. Depending on conceptualobjects captured in the domain model,the words constituting the direct objectphrase are selected; here we acquire theevent token ?SL.submitItemDescriptionto represent the step 1. The step 2describes an action internally performedby the SuD, which we represent with theevent token #validateDescription.

Traditionally, use cases are written in plain English to make the requirementspecifications accessible to a wide audience. It is widely accepted that automaticprocessing of textual use cases is not feasible in general; this is considered asa trade-off for the high readability of textual use cases. Nonetheless, there isongoing research exploring the options to apply linguistic techniques to textualuse cases. The Procasor project of the Distributed Systems Research Group atthe Charles University in Prague aims to convert textual use cases into event-based behavior specifications, behavior protocols.

From Textual Use Cases to Behaviour Specificationsby Vladimir Mencl

P2P communication paradigms arequickly gaining in popularity due to theirscalability and self-organizing nature,forming the basis for building large-scale similarity search indexes at lowcost. However, most of the numerousP2P search techniques proposed inrecent years have focused on single-keyretrieval. A good example is the ContentAddressable Network (CAN), which is adistributed hash table abstraction overCartesian space.

Our objective is to develop a distributedstorage structure for similarity search inmetric spaces that would scale up with(nearly) constant search time. Theadvantage of the metric space approachfor data searching is its 'extensibility',allowing us to perform exact match,range, and similarity queries on anycollection of metric objects. Since any

vector space is covered by a metric spacewith an appropriate distance function(for example the Euclidean distance),even n-dimensional vector spaces arehandled easily. Furthermore, there arenumerous metric functions able to quan-tify similarity between complex objects,such as free text or multi-media objectfeatures that are very difficult to manage.For example, consider the edit distancedefined for sequences and trees, theHausdorff distance applied forcomparing shapes, or the Jacard coeffi-cient, which is often used to assess thesimilarity of sets. Much research is nowfocused on developing techniques tostructure collections of metric objects sothat search requests can be performedefficiently.

A convenient way to assess similaritybetween two objects is to apply metric

functions to decide the closeness ofobjects as a distance, which can be seenas a measure of the objects 'dis-simi-larity'. For any distinct metric objects x,y, z, the distance must satisfy the proper-ties of reflexivity, d(x,x) = 0, strict posi-tiveness, d(x,y) > 0, symmetry, d(x,y) =d(y,x), and triangle inequality, d(x, y) ≤d(x, z) + d(z, y).

The distributed environment iscomposed of network nodes (peers),which hold metric objects, execute simi-larity queries on stored data and commu-nicate with other peers. Every peer isuniquely denoted by its identifier PID.Peers hold data in a set of buckets. Eachbucket has a unique identifier within apeer, designated as BID. Each peer alsomaintains a tree structure called theaddress search tree (AST). This is usedto route similarity range queries through

GHT*: A Peer-to-Peer System for Metric Databy Michal Batko, Claudio Gennaro and Pavel Zezula

GHT* is a scalable and distributed similarity search structure that has beenspecifically designed to support metric space objects. The structure is based onthe Peer-to-Peer (P2P) communication paradigm. It is scalable, which means thatthe query execution achieves practically constant search time for similarity rangequeries in data-sets of arbitrary size. Updates are performed locally and nodesplitting does not require sending multiple messages to many peers.

65



In a subsequent stage, we convert thewhole use case into a behavior protocol(IEEE Trans. Software Eng. 28(11),developed within our group), a formalmethod featuring, eg, a decidable compli-ance relation. We have implemented theconversion outlined here in the Procasortool (Pro-case stands for Protocol usecase). As the tool is employing a statis-

tical natural language parser, the output isonly an estimate of the behavior specifi-cation; nonetheless, it can prove useful inobtaining an initial design of interfaces ofthe future system. Our future workfocuses on enhancing the transformationto produce matching event tokens forcomplementary actions; the long-termgoal is to apply reasoning available for

behavior protocols to detect inconsisten-cies in use case specifications.

Link:http://nenya.ms.mff.cuni.cz/

Please contact:Vladimir Mencl, Charles University Prague/CRCIM,Czech RepublicTel: +420 2 2191 4232E-mail: [email protected]

Figure 1: Fragment of a textual use case.

Use Case: #1 Seller submits an offer

Scope: Marketplace

SuD: Marketplace Information System

Actors: Seller, Trade Commission

Main success scenario specification:

1. Seller submits item description.

2. System validates the description.

...

Figure 2: Parse tree of sentence "Seller submits item description".



the distributed network. AST is based onthe metric data partitioning principleknown in the literature as theGeneralized Hyperplane Tree (GHT).GHT is a binary tree with metric objectsstored in leaf nodes implemented asbuckets of fixed capacity. Inner treenodes contain selected pairs of objectscalled pivots. Respecting the metric, theobjects closer to the first pivot appear inthe left subtree and those closer to thesecond are in the right subtree.

We start by building a tree with only oneroot node represented by a bucket B0When the bucket B0 is full we must splitit: we create a new empty bucket B1 andmove some objects (half of them ifpossible) into it to gain space in B0. SeeFigure 1. The split is done by choosing apair of pivots p1 and p2 from B0 andmoving into bucket B1 all objects o thatsatisfy the condition d(p1, o) > d(p2, o).

Pivots p1 and p2 are associated with anew root node and thus the tree growsone level. This split algorithm can beapplied on any leaf node and is anautonomous operation (no other treenodes need to be modified).

The most important advantage of GHT*with respect to single-site access struc-tures is its scalability through paral-lelism. As the size of a data-set grows,new server nodes are plugged in andboth their storage and the computingcapacity are exploited. Figure 2 showsthe result of the parallel response time ofa range query, determined as themaximum of the costs incurred onservers involved in the query plus thesearch costs of the AST. For evaluationpurposes, we use the number of distancecomputations (both in the AST and in allthe accessed buckets) as the computa-tional costs of a query execution. Thisexperiment shows how the parallelsearch time becomes practically constantfor arbitrary data volume and the largerthe dataset the higher the potential forinterquery parallelism. This is the mostimportant feature of GHT*.

Our future work will concentrate onstrategies for updates (object deletion)and pre-splitting policies, plus moresophisticated mechanisms for organizingbuckets.

Please contact:Claudio Gennaro, ISTI-CNR, ItalyTel: +39 050 315 2888E-mail: [email protected]

Figure 1: Split of a GHT bucket.

Figure 2:

Parallel response time of a range query.

Many research groups at CWI share aneed for high-performance computerfacilities. In the FACS (Facilitating theAdvancement of ComputationalScience) project, three of these groupscombined their resources to set up alarge shared cluster. Since each usergroup has its own specific demands onmemory, CPU power and bandwidth, theFACS cluster deviates from the moreusual symmetric design. The cluster's

edge serves as a compute farm. Itcontains several groups of machines,whose main task is to provide CPUcycles. The core consists of a smallnumber of machines, connected by ahigh-bandwidth low-latency network(Infiniband) and each equipped with alarge amount of memory. Currently, wehave 22 (mostly dual CPU) machines inthe edge and two quad AMD Opteronswith 16G each in the core.

The Database Architectures andInformation Access group works ondatabase technology and needs machineswith both as much memory as possibleand huge, fast storage. For this type ofwork, the bandwidth between memoryand storage is critical. Furthermore, partof the research consists of designinghighly efficient systems for extremelylarge databases. Since the researchconsiders typical database hardware set-

The Centre for Mathematics and Computer Science (CWI) performs a wide rangeof research. Consequently, its high-performance computing facilities must meetdiverse requirements. This has led to the purchase of a cluster with an asymmetriccore-edge design. Under the influence of this asymmetric structure, some clusterapplications are evolving into Grid applications.

CWI's FACS Cluster - Facilitating theAdvancement of Computational Scienceby Stefan Blom, Sander Bohte and Niels Nes



ups, the large-memory fast-disk multi-CPU nodes of the cluster core are a goodmodel for such set-ups.

The Evolutionary Systems and AppliedAlgorithmics group is working onmachine-learning algorithms and agenttechnology. The main issue here is suffi-cient CPU cycles for testing variationson adaptive algorithms and for designingand simulating (large) multi-agentsystems.

The third group, Specification andAnalysis of Embedded Systems, workson explicit-state model checking. Inmodel checking the task is to verify thata model of a system satisfies a property.This can be split into two steps. The firststep is generating the model (a very largegraph called a state space), which is aCPU-intensive task; the second step ischecking if the desired property holds inthe model, which is a memory-intensivetask.

Originally, CWI's model-checkingsystem was set up as a cluster applica-tion. Both steps were performed on the

same types of machines. The choice ofan asymmetric cluster has forced theEmbedded Systems group to modify thesystem to a more Grid-like application,which assigns a task to the part of thenetwork best equipped to handle it.

To achieve this, the verification step isseparated into two stages. In the firststage, the state space is reduced modulobisimulation. The second stage verifiesthe property on the reduced state space.The reduction step yields a state spacethat has the same properties as the orig-inal but is often an order of magnitudesmaller. To exploit this, the EmbeddedSystems group has developed distributedreduction tools.

Without a distributed reduction stage,distributed state-space generation caneasily produce state spaces that do not fitinto the memory of a single machine.This would make validation and reduc-tion impossible. Reduction can be usefuleven when using distributed model-checking tools, as the memory use ofmodel-checking tools grows as a func-tion of not only the size of the state

space, but also the complexity of theproperty.

Using these tools, the work on state-space generation can be focused on theedge part of the cluster, whereas the veri-fication step can take as much advantageas possible from the fast interconnect ofthe core. At the end of this evolutionarypath lies a true Grid application, in whichstate-space generation is performedwherever there is CPU time available,and verification is performed on either ahigh-performance cluster or a large SMPwithin the grid.

The main issue currently holding back aGrid solution is the amount of data thatmust be transferred. In the model-checking system this can be up to halfthe memory. For database research thebandwidth requirements are even larger.The local storage of clusters and super-computers is usually capable of dealingwith these amounts of data. However,the networks connecting the variousGrid locations are typically quite slowcompared to the internal networks. Thiscan result in hours of data transfer forperhaps ten minutes of computation.DATAGRID research and infrastructureshould allow model checking to work ona grid. Eventually, the internal databaseoperations considered by the databasegroup can become a Grid topic as well,but this a more distant possibility.

The computational tasks of research intomachine learning and agent technologycould be met with Grid solutions.However, the rapid prototyping typicallyinvolved makes a small cluster an effi-cient least-effort solution. As gridsbecomes more pervasive and transparentto use, a transition to Grid technologyishighly probable.

Link:http://db.cwi.nl/projecten/project.php4?prjnr=162

Please contact:Stefan Blom, CWI, The NetherlandsTel: +31 20 592 4280E-mail: [email protected]

Sander Bohte, CWI, The NetherlandsTel: +31 20 592 4074E-mail: [email protected]

Niels Nes, CWI, The NetherlandsTel: +31 20 592 4098E-mail: [email protected]

Applications of the FACS cluster include simulations of multi-agent systems such as

DEAL (Distributed Engine for Advanced Logisitics). In this project, agents are used

to compute efficient schedules for freight trucks.



AAA/SynDEx provides a formal frame-work based on graphs and system-levelCAD software. On the one hand, thesespecify the functions of the applications,the distributed resources in terms ofprocessors and/or specific integratedcircuit and communication media, andthe non-functional requirements such asreal-time performances. On the other,they assist the designer in implementingthe functions onto the resources whilesatisfying timing requirements and, asfar as possible, minimizing theresources. This is achieved through agraphical environment, which allows theuser to explore manually and/or auto-matically the design space solutionsusing optimization heuristics.Exploration is mainly carried outthrough timing analyses and simulations.The results of these predict the real-timebehaviour of the application functionsexecuted on the various resources, ieprocessors, integrated circuits andcommunication media. This approachconforms to the typical hardware/soft-ware codesign process. Finally, for thesoftware part of the application, code isautomatically generated as a dedicatedreal-time executive, or as a configurationfile for a resident real-time operatingsystem such as Osek or RTlinux.

This approach will improve the designsafety provided by formal models, anddecrease the development cycle thanksto timing simulation and automatic codegeneration. Another interesting feature is

the ability to easily interface AAA/SynDEx with domain-oriented specifica-tion languages such as the Synchronouslanguages, AIL for automobile,Scilab/Scicos for automatic control, AVSfor image processing and so on. Thismeans a link can be provided to users’most commonly used tools, and whenthese languages guarantee timing proper-ties, that these will be maintained duringthe distributed real-time implementation.This is particularly important in the caseof safety-critical applications found in thefield of avionics and automobiles.

Figure 1 shows the principles of SynDExinterfaced with the hybrid dynamicsystems modeller and simulator Scicos.

Figure 2 shows the SynDEx graphicaluser interface used to design a manualdriving application with joystick for theCyCab modelled and simulated withScicos. The CyCab is an intelligent andmodular electric vehicle designed atINRIA Rocquencourt by the IMARAteam and industrialized by Robosoft. Itsarchitecture is based on four MPC555microcontrollers and an embedded PC,all of which communicate through aCAN bus.

Since the early nineties, there has been asignificant amount of research onAAA/SynDEx at INRIA Rocquencourt,firstly by the SOSSO, then the OSTRE,and now the AOSTE team. As time has

by Yves Sorel

Distributed real-time embedded systems are of crucial importance in applicationdomains such as avionics, automobiles, telecommunication and mobile robotics.Principally based on digital electronics, which includes software, this aspect ofthe various applications is rapidly growing. Because they must meet constraintson resource distribution and optimization, as well as time, the design of suchsystems is particularly complex. In order to assist the designers, scientists atINRIA are proposing a methodology called AAA (Algorithm ArchitectureAdequation) and its associated system-level CAD software called SynDEx. Thesecover the whole development cycle, from the specification of the applicationfunctions, to their implementation running in real time on a distributed architecturecomposed of processors and specific integrated circuits.

SynDEx: System-Level CAD Software for Optimizing Distributed Real-TimeEmbedded Systems

Figure 1:

SynDEx

interfaced

with Scicos.



passed, it has increased in popularitywith industrial users, and has thereforebeen evaluated by several companies.This occurred firstly in France and threeyears ago in Europe, through ITEAprojects launched in automobile andtelecommunications research. It is alsocurrently used in real-world applications

provided by Robosoft, MBDA,Mitsubishi-ITE, and PSA.

AAA/SynDEx runs under Unix/Linux,Windows and MacOS and comes withfull documentation, including a referencemanual, a user manual and a tutorial. It isdownloadable free of charge under

INRIA copyright athttp://www.syndex.org.

Future work will focus on the followingaspects, from high to low levels of thedesign flow: the integration ofAAA/SynDEx with the MDA (ModelDesign Architecture) approach byproposing successive model transforma-tions; the extension of currently staticoptimization techniques to more dynamicschemes in order to better support eventhandling and dynamic creation of func-tions; and finally, the tight coupling ofsystem-level and circuit-level CAD toolsin order to actually provide automatichardware/software partitioning in the co-design process.

Links: http://www.syndex.orghttp://www.scilab.orghttp://www-rocq.inria.fr/imara,http://www.robosoft.fr

Please contact:Yves Sorel, INRIA, FranceTel: +33 1 39 63 52 60E-mail: [email protected]

Figure 2: SynDEx GUI used to design a manual driving application with a joystick for

the CyCab.

Due to the high structural organization ofthe intracellular space, diffusion isalways restricted to the specific shape ofthe organelle under consideration. Todetermine and quantify the diffusivemobility of substances within live cellsthe technique of fluorescence recoveryafter photobleaching (FRAP) is widelyemployed.

From a practical point of view a methodis needed to deduce molecular (micro-scopic) solute diffusion constants frommeasured fluorescence recovery data.

We propose the deterministic particlemethod of Particle Strength Exchange(PSE) for simulations of diffusion incellular geometries of realisticcomplexity. PSE was introduced as analternative to the method of randomwalk, enabling higher-order simulationsof convection-diffusion processes. PSEis a grid-free, deterministic particlemethod and thus combines the advan-tages of high-order convergence andgeometric flexibility. It enables efficientsimulations of solute diffusion in biolog-

ical structures using the geometry of realsamples.

Due to the complex shape of the compu-tational domain, a large number of parti-cles is usually needed to resolve theconcentration field. We implemented anefficient parallel PSE code based on ageneral-purpose Parallel Particle Mesh(PPM) Library currently under develop-ment. The code provides differentdomain decomposition techniques,dynamic load balancing among inhomo-geneous processor clusters, parallel file

Large-Scale Simulations of Diffusion in Cell Biologyby Ivo F. Sbalzarini, Anna Mezzacasa, Ari Helenius and Petros Koumoutsakos

Molecular transport within the intracellular structures of live biological cells isdominated by diffusion in confined compartments with complex geometries.Quantitative evaluations of standard experimental assays in cell biology (eg tomeasure diffusion coefficients of proteins in vivo) require knowledge of thesolution of the diffusion equation in such geometries. We present a high-performance parallel implementation of a particle method to solve the diffusionequation in 3D reconstructions of real samples obtained by fluorescence confocalmicroscopy.

70



I/O and efficient network communica-tion. It was tested and optimized both ondistributed memory and shared memorysystems with either scalar or vectorprocessors. Arbitrary geometries can beread in as triangulated surfaces and parti-cles are initialized inside it. (Using aparticle method, there is no need for gridgeneration.)

The method was successfully applied todiffusion in the Endoplasmic Reticulum(ER), which is the most prevalentexample of a highly convoluted andinterconnected structure in three dimen-sions. The geometry of several ERsamples was reconstructed by recordingthe shape of the ER samples from livecells using stacks of serial sectionsobtained by confocal fluorescence lightmicroscopy. It was checked that thereconstructed 3D geometry of theorganelle was connected and exhibitedall the topological properties of a realER. It could thus be used directly as acomputational domain. Figure 1 shows3D visualizations of the evolvingconcentration field inside a sample ERgeometry. Using reconstructions fromreal samples allowed direct comparisonsbetween simulation results and experi-mental data from the same ER.

Fitting simulated and experimentallymeasured diffusion curves using timestretching enabled us to determine theunknown molecular diffusion constant inlive cells. By using the same geometryboth for experiment and simulation, allneed for modelling its effect on theapparent diffusion constant is eliminated

and the only parameter left is the molec-ular diffusion constant itself, taking theinfluences of the specific geometry athand into full account.

Please contact: Petros Koumoutsakos, ETH Zurich/SARIT, SwitzerlandE-mail: [email protected]

With the establishment of the IEEE-802.11a/b/g standards, wirelesscomputer networks have started to grad-ually replace regular wired Internetaccess in homes, offices, and publicareas, such as train stations and airports.However, as only limited bandwidth isallocated to be used by such systems,data rates need to be shared. As thenumber of users grows and applications

become more demanding in terms ofthroughput and link reliability (quality ofservice), current standards can no longermeet the requirements.

Multiantenna (MIMO) systems are ameans to increase the data rate, link reli-ability, and range of all kinds of wirelessnetworks and mobile communicationsystems, without increasing bandwidth

and/or transmit power. Three maineffects are responsible for this perfor-mance gain: The 'diversity gain' stabi-lizes the link, as the additional antennascan be used to compensate for a weakconnection on any of the other antennas.The 'array gain' allows the receiver topick up more of the energy that was radi-ated from the transmitter, effectivelyincreasing the range of the system. The

At ETH Zurich, a MIMO-WLAN (Multiple-Input Multiple-Output Wireless Local AreaNetwork) system has been developed and demonstrated successfully in a testsetup. Four antennas at the transmitter and at the receiver boost the data ratefrom 54 million bits per second (Mbps) in a regular 802.11g-based WLAN to upto 216 Mbps. In addition to the higher throughput, the system provides increasedlink reliability and better coverage.

Real-Time MIMO Testbed for Next Generation Wireless LANs by Andreas Burg and Helmut Bölcskei

Snapshots of the concentration distribution from a sample PSE simulation. The ER

membrane is visualized as a transparent surface, and the concentration of the protein

under consideration as a volume density cloud inside it. The initial concentration field

was 1 everywhere outside the cubic box shown, and 0 inside it. The region of interest

around the box is enlarged.



'multiplexing gain' finally allows trans-mitting multiple data streams concur-rently in the same frequency band,thereby effectively increasing the datarate. A schematic of such a system withthree antennas is shown in Figure 1. Thehigh-rate data stream is first partitionedinto three lower rate streams, which arethen sent simultaneously from the indi-vidual antennas, using the samefrequency band. At the receiver, threeantennas pick up the superposition of thestreams, and a suitable decoder separatesthem again and reconstructs the originalhigh-rate stream.

Researchers at the CommunicationTechnology Laboratory and at theIntegrated Systems Laboratory of the

Swiss Federal Institute of Technology(ETH) Zurich are working on the devel-opment of such MIMO (Multiple-InputMultiple-Output) systems and on theirintegration in today's WLAN systems.

Research activities focus on both thetheoretical analysis and the VLSI imple-mentation of such systems. At theCommunication TechnologyLaboratory, the limits of multiantennasystems are investigated, and new algo-rithms that approach these limits in prac-tical systems are developed. The addi-tional degrees of freedom induced bymultiple antennas can be exploited bydifferent algorithms, which need to betailored to individual applicationscenarios. The implementation

complexity of these algo-rithms generally exceedsthe complexity of algo-rithms applied in single-antenna systems signifi-cantly. For this reason,researchers at the IntegratedSystems Laboratory areconcerned with algorithmoptimization and efficientimplementation in applica-tion-specific integratedcircuits (ASIC). The chal-lenge is to meet throughputrequirements while main-taining a small chip area toreduce cost and to guaranteelow power consumption. Arecent example of this jointeffort is the development of

a decoder ASIC (sphere decoder) thatallows the optimum separation ofparallel data streams, received onmultiple antennas.

Within the scope of a three years project,financed by the ETH research council, areal-time demonstrator for a MIMO-WLAN system with four antennas eachat the transmitter and the receiver hasbeen developed. The system boosts thedata rate of a current single antennaIEEE-802.11a WLAN system to 216Mbps, without the need for additionalbandwidth and/or transmit power. Suchinitial practical experiments allowresearchers to test new algorithms inreal-time and in real-world scenarios.They are also indispensable contribu-tions to standardization efforts (eg, theongoing work in the high-throughput802.11n working group) to identifyimplementation bottlenecks and otherpitfalls that might appear in their hard-ware realization. Recently, first over-the-air experiments have been carriedout, indicating that the predicted gainsare in fact achievable and that theproposed system can be realized withmanageable hardware complexity thatwill eventually allow integration into alaptop computer or personal digitalassistant (PDA).

Please contact:Helmut Bölcskei, ETH Zurich/SARITE-mail: [email protected]

Figure 1: MEMO scheme.

Figure 2: MIMO WLAN system experimental setup — signal processing, MIMo test

equipment, channel characterization.


EVENTS

The main objectives of the Cross-Language Evaluation Forum (CLEF)are to stimulate the development ofmono- and multilingual informationretrieval systems for Europeanlanguages and to contribute to thebuilding of a research community in themultidisciplinary area of multilingualinformation access. These objectives arerealised through the organisation ofannual evaluation campaigns and work-shops. Each campaign offers a series ofevaluation tracks designed to test

different aspects of mono- and cross-language system performance.

One of the principal objectives whenCLEF began was to encourage devel-opers to build multilingual retrievalsystems capable of searching overcollections in a number of languages.The multilingual track was thus themain track for several years and wasmade progressively more difficult. WithCLEF 2003, where the track included atask which involved finding relevant

documents in a collection containingdocuments in eight languages, we feltthat we had achieved an important goal.We had shown that fully multilingualretrieval could be (almost) as effectiveas bilingual (L1 -> L2) retrieval and thatsystems are able to adapt and reengineerrapidly and effectively to process newlanguages as the need arises. For thisreason, in CLEF 2004 we decided toreduce the multilingual documentretrieval activity to leave more space for

by Carol Peters

The results of the fifth campaign of the Cross-Language Evaluation Forum werepresented at a two-and-a-half day workshop held in Bath, UK, 15-17 September,immediately following the eighth European Conference on Digital Libraries. Theworkshop was attended by nearly 100 researchers and system developers.

Cross-Language Evaluation Forum — CLEF 2004

The conference included, from August22 to 24, 2004, two days and a half ofworkshops (27 workshops) and tutorials(11 tutorials) on most advanced topics,as well as the second starting AIResearchers Symposium (STAIRS),followed by the three-day technicalprogram of ECAI. Following an estab-lished tradition, the sub-conference onprestigious applications of intelligentsystems (PAIS) ran in parallel withECAI. The technical programmeincluded refereed paper presentations,invited talks by prestigious speakers andposter sessions. The invited speakerswere Glorianna Davenport (MIT, USA),

Christian Freksa (University of Bremen,Germany), Carole Goble (University ofManchester, UK), and Seppo Laukkanen(SenseTrix, Finland).

Among 653 submissions received from45 countries, 168 papers were acceptedfor oral presentation at ECAI 2004, 13for oral presentation at PAIS 2004, and,for the first time, 87 were accepted asposters.

Although the majority of the submittedpapers were from Europe (505), therehas been a significant number of papersfrom the rest of the world, and particu-

larly from Asia (53), North America (50)and Oceania (21). We can also notice theemergence of high quality AI research inSouth America (16) and Africa (8), andthis is an indicator of the increasinginterest in AI worldwide.

After the late registration deadline themain conference (ECAI’04) has 694registrations, STAIRS has 53 registra-tions, and 436 people have registered forthe 27 workshops.

More information: https://www.dsic.upv.es/ecai2004/

by Vicent J. Botti

The 16th European Conference on Artificial Intelligence (ECAI 2004) was held inValencia, Spain, from 22-27 August 2004, hosted by the Universidad Politécnicade Valencia and the Grupo de Investigación de Tecnología Informática –InteligenciaArtificial (GTI-IA). ECAI provided a public forum for researchers from academyand industry and public organizations, offering them an invaluable occasion tomeet and exchange ideas, thus contributing to making real the upcoming'Information Society'.

16th European Conference on Artificial Intelligence


EVENTS

other types of cross-language informa-tion retrieval experiments.

Six tracks were offered to evaluate theperformance of systems for:• mono-, bi- and multilingual document

retrieval on news collections (Ad-hoc)• mono- and cross-language domain-

specific retrieval (GIRT)• interactive cross-language retrieval

(iCLEF)• multiple language question answering

(QA@CLEF)• cross-language retrieval on image

collections (ImageCLEF)• cross-language spoken document

retrieval (CL-SDR).

CLEF 2004 thus marked a breakingpoint with respect to previouscampaigns. The focus was no longerconcentrated on multilingual documentretrieval but was diversified to includedifferent kinds of text retrieval acrosslanguages (exact answers in the ques-tion-answering track) and retrieval ondifferent kinds of media (not just plaintext but collections containing imageand speech as well). In addition,increasing attention was given to issuesthat regard system usability and usersatisfaction with tasks to measure theeffectiveness of interactive systems orsystem components being included inboth the QA and the ImageCLEF trackswith the collaboration of the coordina-tors of iCLEF (the interactive track).

In order to cover all these activities, theCLEF test collection has been expandedconsiderably: the main multilingualcomparable corpus now contains almost2 million news documents in tenlanguages. A secondary collection usedto test domain-specific system perfor-mance consists of the GIRT-4 collectionof English and German social sciencedocuments. ImageCLEF used twodistinct collections: a collection ofhistoric photographs provided by StAndrews University, Scotland, and acollection of medical images withFrench and English case notes madeavailable by the University Hospitals,Geneva. Finally, the cross-languagespoken document retrieval track (CL-SDR) used speech transcriptions inEnglish from the TREC-8 and TREC-9SDR tracks, supplied by the National

Institute of Standards and Technology(NIST), USA.

The response from the informationretrieval community was very encour-aging. Participation in this year’scampaign was considerably up withrespect to the previous year with 55groups submitting results for one ormore of the different tracks: 36 fromEurope, 13 from N.America; 4 fromAsia and one mixed European/Asiangroup. As in previous years, participantsconsisted of a nice mix of new-comersand veteran groups. The success of thequestion answering and image retrievaltracks had a big impact on participationin CLEF 2004, not just with respect tothe numbers but also regarding the skills

and expertise involved. The popularityof question answering has meant that agrowing number of participants have anatural language processing backgroundwhile the image and spoken documentretrieval tasks have brought in groupswith experience in diverse areasincluding speech recognition, imageprocessing and medical informatics –making CLEF an increasingly multidis-ciplinary forum.

The campaign culminated in the work-shop held in Bath, UK, 15-17September. In addition to presentationsby participants in the campaign, talksincluded reports on the activities of the

NTCIR evaluation initiative for Asianlanguages, and on industrial experiencein building cross-language applications.The final session consisted of a panel inwhich panellists attempted to analysethe current organisation of the CLEFcampaigns in depth, consideringwhether we are working on the rightproblems, choosing our investmentswisely, and giving sufficient attention tothe user perspective. Tracks taken intoconsideration for the CLEF 2005campaign include multilingual Webretrieval and a cross-languageGeographic Information Retrieval track.

The presentations given at the CLEFWorkshops and detailed reports on theexperiments of CLEF 2004 and previous

years can be found on the CLEF websiteat http://www.clef-campaign.org/

CLEF is an activity of the DELOSNetwork of Excellence for DigitalLibraries.

Link:http://www.clef-campaign.org

Please contact:Carol Peters, ISTI-CNR, ItalyCLEF CoordinatorTel: +39 050 3152897E-mail: [email protected]

CLEF workshop participants.


EVENTS

CALL FOR PAPERS

CAiSE 2005 — The 17thConference on AdvancedInformation SystemsEngineering

Porto, Portugal, 13-17 June 2005

Since the late 1980's, the CAiSE confer-ences provide a forum for the presenta-tion and exchange of research results andpractical experiences within the field ofInformation Systems Engineering.CAiSE’05 aims at bringing togetherresearchers, users, and practitioners inthe field of information systems engi-neering. The conference programme willfeature paper presentations, workshops,tutorials, and interactive panel sessions.

Conference Theme The Internet has been changing thesociety and the economy, and the wayinstitutions and businesses operate hasevolved rapidly. All citizens are nowexpected to interact directly with tech-nology based systems without directhuman intermediation, using differentinterface appliances. Moreover, manyinformation systems interact with eachother with limited human supervision.The Internet and its infrastructure play acritical role in this evolution, and it is ofthe utmost importance that semanticaspects are taken into consideration on asound basis. Improved communicationand understanding between people, thefinal objective of advanced informationsystems, requires improved communica-tion and understanding between systemsand its developers. Advanced informa-tion systems engineering therefore needsto keep improving its methods and tech-nologies in order to support the properdevelopment of the economy and of oursociety.

Relevant Topics In addition to the special theme, topicsrelevant for submissions to CAiSE 2005include, but are not limited to, thefollowing:

• Methodologies, Models, and Tools forIS Development

• Requirements Engineering for IS • Model Driven Architectures

• Enterprise Modelling Methods andTools

• Service Oriented and MobileComputing

• E-government Strategies and Applica-tions

• Knowledge, Information and DataQuality

• Web Content Management andDistribution

• Workflow Systems • Knowledge Management • Model and Software Reusability • Data Warehousing & OLAP • Metadata and Ontologies • Support for Co-operative Work • Innovative Database Technologies • Distributed, Mobile, and Open

Architectures • Agent-Oriented Technologies • Languages and Protocols for IS • Component-Ware and IS • IS Reengineering • IS Usability and Interfaces to IS • Simulation in IS Development • Semantic Web.

Case studies and experience reportsrelated to the above topics are alsowelcome

Important Dates • Paper Submission: 30 November 2004• Notification of Acceptance: 12

February 2005• Final Paper Submission: 12 March

2005

The conference is organized by FEUP -Faculdade de Engenharia daUniversidade do Porto (www.fe.up.pt)and DSIC-UPV - Departament deSistemes Informàtics i Computació,Universitat Politècnica de València(www.dsic.upv.es) and sponsored byERCIM.

More Information http://www.fe.up.pt/caise2005/

CALL FOR PAPERS

Mirage 2005

Rocquencourt (near Versailles),France, 1-2 March 2005

MIRAGE 2005 is an internationalconference with focus on ComputerVision/Computer Graphics collaborationtechniques involving image analysis/synthesis approaches.

In the domain of Computer Vision, thisanalysis by synthesis collaboration maytake the form of model-based approachesfor which the analysis, recognition orunderstanding process does not rely onthe data only but is linked to the opti-mization of a parametric model intro-duced A PRIORI. In the field ofComputer Graphics, this thematic refersto image-based approaches which foundthe modeling, animation, lighting orrendering process on the analysis of realpre-existing data in order to gain inproductivity or in realism. In both cases,this extraction and synthesis collabora-tion can be iterative and conducted in aloop: the synthesized information layingthe foundations for a better characteriza-tion that will in turn enable the synthesisstep. Augmented Reality is a directapplition of this collaboration .

Authors are encouraged to submit paperson theoretical, computational, experi-mental or industrial aspects of model-based image analysis and image-basedmodel synthesis.

Topics Topics include (but are not limited to) :• model-based imaging and analysis• iImage-based modeling and 3D recon-

struction• data driven animation• image and video-based lighting and

rendering• model-based vision approaches• model-based indexing and database

retrieval• model-based object tracking in image

sequences• model-based image and shape analysis• model-based video compression tech-

niques.

With applications in the field of :• human/Computer interfaces


EVENTS

CALL FOR PAPERS

HCI International 2005

Las Vegas, Nevada, USA, 22-27 July 2005

The 11th International Conference onHuman-Computer Interaction is jointlyheld under one management and oneregistration with the Symposium onHuman Interface (Japan) 2005, the 6thInternational Conference on EngineeringPsychology & Cognitive Ergonomics,the 3rd International Conference onUniversal Access in Human-ComputerInteraction, the 1st InternationalConference on Virtual Reality, the 1stInternational Conference on Usabilityand Internationalization, the 1stInternational Conference on OnlineCommunities and Social Computing,and the 1st International Conference onAugmented Cognition.

HCI International 2005 and the affiliatedconferences will be held at the CaesarsPalace Hotel, one of the most prestigiousresorts in the world, under the auspicesof 9 distinguished international boards of202 members from 31 countries!

The conference objective is to provide aninternational forum for the disseminationand exchange of scientific informationon theoretical, generic, and applied areasof HCI, usability, internationalization,virtual reality, universal access andcognitive ergonomics. This will beaccomplished through the following sixmodes of communication: plenarypresentation, parallel sessions, demon-stration and poster sessions, tutorials,exhibitions and meetings of specialinterest groups. The six-day conferencewill start with three days of tutorials. Thetutorials will begin on Friday, 22 July2005. A total of 18 tutorials will beoffered (both half-day and full-day) atintroductory, intermediate, andadvanced levels covering the entire spec-trum of the 9 tracks and 7 conferences.

HCI International conference gathersover a thousand scientists, academics andprofessionals from research institutions,universities, and companies from all overthe world. The HCI International 2005exhibits will provide state-of-the-art HCI,VR, psychology and computer products

and services for the users, professionals,and researchers in the field. Attendeeswill be able to examine state-of-the-artHCI technology and interact with manu-facturing representatives, vendors,publishers, and potential employers.Exhibitors who commit to exhibit byNovember 20, 2004, apart from the factthat they will have a better opportunity inbooth location selection and a very attrac-tive pricing, they will be listed in theAdvance Program, Final Program, andthe Conference Proceedings, which willbe distributed worldwide. Exhibitors whocommit to exhibit by May 1, 2005 will belisted in the Final Program.

Thematic Areas• Ergonomics and Health Aspects of

Work with Computers;Program Chair: Pascale Carayon

• Human Interface and the Managementof Information;Program Chair: Michael J. Smith

• Human-Computer Interaction;Program Chair: Julie Jacko

• Engineering Psychology and CognitiveErgonomics;Program Chair: Don Harris

• Universal Access in Human-ComputerInteraction;Program Chair: Constantine Stephanidis

• Virtual Reality;Program Chairs: Kay Stanney &Michael Zyda

• Usability and Internationalization;Program Chair: Nuray Aykin

• Online Communities and SocialComputing;Program Chair: Jennifer Preece

• Augmented Cognition;Program Chair: Dylan D. Schmorrow.

Important Dates• Paper Presentations - Deadline for

Abstract Receipt: 1 October 2004• Posters/demonstrations - Deadline for

Abstract Receipt: 1 April 2005• Special Interest Groups - Deadline for

Abstract Receipt: 20 October 2004• Tutorials - Deadline for Abstract

Receipt: 20 October 2004• Early Registration: 20 January 2005 • Mid Registration: 21 January - 20 May

2005 • Late Registration: After 20 May 2005

More information:http://www.hci-international.org/

CALL FOR PARTICIPATION

EGC2005 — European Grid Conference 2005

Amsterdam, The Netherlands, 14-16 February 2005

The European Grid ConferenceEGC2005 is the premier event on GridComputing in Europe. It will focus on allaspects of Grid computing in Europe anda such will bring together participantsfrom research and industry. EGC 2005 isa follow-up of the AcrossGridsConferences held in Santiago deCompostela (2003) and in Nicosia(2004).

EGC2005 will have three main tracks:• a scientific track;• an industrial track;• a special events track.

Together they aim to cover the completefield of Grid developments in Europe.

More information: http://genias.biz/egc2005

• video-games and entertainmentindustry

• mMedia production from film, broad-cast and games

• post-production, computer animation,virtual special effects

• realistic 3D simulation , virtual proto-typing

• multimedia applications, multimediadatabases classification

• virtual and augmented reality• medical and biomedical applications.

Important Dates• 15 November 15, 2004: Full paper

submission• 20 December 2004: Notification of

acceptance• 15 January 2005: Camera-ready

papers due, early registration deadline• 1 February 2005: Registration dead-

line for authors of accepted papers• 15 February 2005: Late registration

deadline• 1-2 March 2005.

More information: http://acivs.org/mirage2005/


EVENTS

CALL FOR PAPERS

ECDL 2005 — EuropeanConference on Digital Libraries

Vienna, Austria, 18-23 September 2005

ECDL 2005 is the 9th conference in theseries of European Digital Libraryconferences. ECDL has become themajor European conference on digitallibraries, and associated technical, prac-tical, and social issues, bringing togetherresearchers, developers, contentproviders and users in the field. ECDL2005 is jointly organized by the ViennaUniversity of Technology (VUT), theAustrian National Library (ÖNB), andthe Austrian Computer Society (OCG).The conference will take place in theVienna Technical University.

Important Dates:• workshop proposal deadline:

28 January 2005• workshop acceptance notification:

27 February 2005• paper/tutorial/panel submission dead-

line: 1 March 2005• acceptance notifications: 15 May 2005• final version of papers: 3June 2005.

Topics:Topics of contributions include (but arenot limited to):• concepts of digital libraries and digital

documents• system architectures, integration and

interoperability• information organization, search and

usage• user studies and system evaluation• digital preservation• digital library applications.

More information:http://www.ecdl2005.org

CALL FOR PARTICIPATION

Virtual Reality 2005

Bonn, Germany, 12-16 March 2005

IEEE VR 2005 is the premier interna-tional conference and exhibition invirtual reality, and provides a trulyunique opportunity to interact withleading experts in VR and closely relatedfields such as augmented reality, mixedreality and 3D interaction. Test your ownwork and educate yourself through expo-sure to the research of your peers fromaround the world. And of course, therewill be time to renew friendships, makenew ones and experience the Europeancuisine and culture in Bonn, the formerGerman capital.

IEEE VR is an excellent opportunity topresent:• leading edge research contributions

(papers)• innovative applications and experi-

ence reports (application sketches)• communicate the state of the art

(survey reports)• present and discuss solutions (panels)• introduce to your VR research (lab

presentations).

The main conference starts with tutorialsand workshops on March 12 and 13.They will introduce to specific topics ofthe field and give a compact view ofongoing research activities.

The industrial exhibition and theresearch lab demonstrations will providepractical insights into solutions andprototypes.

IEEE VR2005 is cohosted by:• caesar: center for advanced European

studies and research• University of Applied Sciences Bonn-

Rhein-Sieg• Fraunhofer Institute for Applied

Information Technology• Rheinische Friedrich-Wilhelms-

Universitaet Bonn.

The conference is sponsored by ERCIM

More information: http://www.vr2005.org

CALL FOR PAPERS

INTERACT 2005 — Tenth IFIPTC13 International Conferenceon Human-Computer InteractionRome, Italy, 12-16 September 2005

The Interact '05 conference will highlightto both the academic and industrial worldthe importance of the Human-ComputerInteraction area and its most recentbreakthroughs on current applications.

Topics:Suggested topics include but are notlimited to:• Multi-modal Interfaces• Context-dependent Systems• End User Development• Intelligent Environments• Tangible Interfaces• Novel User Interfaces• Usability Evaluation• Location-aware Interaction• Model-based Design of Interactive

Systems• Multi-user Interaction• Safety Issues and Human error• Cognitive Models• Visualisation Techniques• Web Design and Usability• Mobile and Ubiquitous Computing• 3D and Virtual Reality Interfaces• Adaptive Interfaces• Web Accessibility• Usability & Accessibility.

Important Dates:Full papers, tutorials, workshops,doctoral consortium:• Submission deadline: 31 January 2005• Notification of acceptance: 15 March

2005• Final version due: 30 April 2005.

Short papers, special interest groups,panels, posters, demos:• Submission deadline: 30 March 2005• Notification of acceptance: 10 May 2005• Final version due: 30 May 2005.

The conference is sponsored by ERCIM.

More information: http://www.interact2005.org/


EVENTS

The International Telecommunication Union (ITU) has just (September 2004)released the sixth book in its Internet Reports series entitled "The PortableInternet". The book examines the possible impact of high-speed wireless

communication used together with portableterminals. The conclusion is that this combinationcould well be a "disruptive technology" with widesweeping impact on global communication.

The book is nicely self contained with a glossary ofterms included. This will be important for somereaders when the myriad of available technologies isdiscussed in the second chapter.

The third chapter deals with market trends. Here, forexample, it is interesting to see that while almost halfof the world's broadband subscribers is already inthe Asia-Pacif ic region, the potential here isenormous. Future broadband development is highlydependent on national policy and regulation, the

subject of the fourth chapter. Case histories show dramatic results ofderegulation. The impact that the portable Internet might have on the “digitaldivide” is covered in the fifth chapter. This is one of the world's largest problemsbut as the cost of the portable Internet decreases perhaps the enormouspotential of the Internet can be realized in developing countries. Future marketapplications are covered in the next chapter. Existing applications in Korea letmobile telephone users pay for many items and services using their mobiletelephone — examples are food and transportation. The last chapter looks at thepotential impact on society and includes mention of the medical concerns thatsome people have for widespread wireless use.

There is an Annex at the end of the book which gives all kinds of statisticalinformation for some 200 countries around the world. Penetration of the Internet,wide band access and mobile telephony, as well as costs, are covered. Suchstatistics could be rather dry but these are so interesting that it is easy to studythem for quite some time. For example, the country with the highest broadbandpenetration is the Republic of Korea with somewhat more than 23%.

The book, over 200 pages, may be purchased via the ITU Web site:http://www.itu.int/osg/spu or http://www.itu.int/publications/bookshop orhttp://www.itu.int/portableinternet/

Harry Rudin, Consultant and Swiss Local Editor for ERCIM News

BOOK REVIEW

SPONSORED BY ERCIM

Workshop on Constructionand Analysis of Safe, Secureand Interoperable SmartDevices

Nice, France, 7-11 March 2005

The CASSIS International Workshopwill be held in the week of March 7th toMarch 11th 2005 in Nice and will consistof keynote and invited presentations byleading academics and industrials in thefield of security and analysis of smartdevices.

The workshop will focus on:• Research trends in smart devices• Web services• Virtual machine technology• Security• Proof-Carrying Code• Validation and formal methods• Embedded devices.

The workshop is organized by theINRIA's Everest team.

More information:http://www-sop.inria.fr/everest/events/cassis05/

CALL FOR PAPERS

ECIR'05 — 27th EuropeanConference on InformationRetrievalSantiago de Compostela, Spain,21-23 March 2005

The annual European Conference onInformation Retrieval Research, orga-nized by the Information RetrievalSpecialist Group of the British ComputerSociety (BCS-IRSG), is the mainEuropean forum for the presentation ofnew research results in the field of infor-mation retrieval.

The organisers encourage the submis-sion of high-quality research papersreporting new, unpublished, and innova-tive research results within informationretrieval. Poster submissions addressingany of the areas identified in the confer-ence topics are also invited. Authors areencouraged to demonstrate work inprogress and late-breaking researchresults. The conference intends to coverall aspects of accessing digital informa-tion without explicitly specified seman-tics. Papers whose sole or main author isan MSc, PhD or postdoctoral student areespecially welcome.

TopicsTopics of interest include but are notlimited to the following areas:• IR models, techniques, and algorithms• Users, society, and IR• IR applications• IR system architectures• Content representation, and

processing• Test and evaluation methods for IR

systems• Multimedia and cross-media IR.

Important Dates• 1 November 2004: paper submission• 8 November 2004: poster submission• 9 December 2004: notification for all

submissions• 22 December 2004: Final copy due.

The conference is sponsored by ERCIM.

More information: http://www-gsi.dec.usc.es/ecir05/


ERCIM News is the magazine of ERCIM.Published quarterly, the newsletter reports onjoint actions of the ERCIM partners, and aimsto reflect the contribution made by ERCIM tothe European Community in InformationTechnology. Through short articles and newsitems, it provides a forum for the exchange ofinformation between the institutes and alsowith the wider scientific community. This issuehas a circulation of over 9000 copies.

Copyright NoticeAll authors, as identified in each article, retaincopyright of their work.

ERCIM News online edition is available athttp://www.ercim.org/publication/ERCIM_News/

ERCIM News is published by ERCIM EEIG, BP 93, F-06902 Sophia-Antipolis CedexTel: +33 4 9238 5010, E-mail: [email protected] 0926-4981

Director: Michel Cosnard

Central Editor:Peter Kunz ([email protected])

Local Editors:AARIT: n.a.

CCLRC: Martin Prime [email protected]

CRCIM: Michal [email protected]

CWI: Annette Kik [email protected]

CNR: Carol Peters [email protected]

FORTH: Eleftheria [email protected]>

Fraunhofer ICT Group: Michael [email protected]

FNR: Patrik [email protected]

FWO/FNRS: Benoit [email protected]

INRIA: Bernard [email protected]

Irish Universities Consortium:Ray [email protected]

NTNU: Truls [email protected]

SARIT: Harry [email protected]

SICS: Kersti [email protected]

SpaRCIM: Salvador [email protected]

SRCIM: Gabriela Andrejkova [email protected]

SZTAKI: Erzsébet Csuhaj-Varjú[email protected]

VTT: Pia-Maria Linden-Linna [email protected]

W3C: Marie-Claire [email protected]

Free subscriptionYou can subscribe to ERCIM News free ofcharge by: • sending e-mail to your local editor• contacting the ERCIM office (see address

above) • filling out the form at the ERCIM website at

http://www.ercim.org/

On the Impact of Broadband InternetFor most European internet users in2000, broadband internet was anunknown form of access.

Fast internet did exist, but it was almostexclusively used by companies anduniversities. Only consumers (akapunters) with cable internet could surfthe net at speeds well over 128Kbps(ISDN). However, due to the specificarchitecture of coax-internet, this kind ofaccess was not common in Europe round2000.

Some reports, mainly based on researchdone in the US, were not very positiveabout the growth perspectives fortelecom operators of Broadband. Therewere two main reason for this:• one renowned research company

briefed to its customers : "Broadbandis perceived as a luxury; until serviceproviders make it a necessity,consumers' interest will remainlimited".

• as most European punters had learneda few years before that there wassomething like 'free access' not manythose days were willing to pay monthlyfees over Euro 50.

Nevertheless during 2002/2003 all overEurope the number of broadband userssurged. More Cable television cable-networks were upgraded to make internetaccess possible. As a result, cableproviders in countries like Belgium,Austria and the Netherlands becameleaders in broadband and in almost everyEU-member state DSL was no longer aservice offered exclusively by the incum-bent telecom operator.

The number of broadband suppliers grewand all were eager to introduce newaccess forms to increase their marketshare. As a result of this in all EUmember states the monthly fees forbroadband access dropped.

So the number of users grew, but why?Was growth only caused by lowerprices?

One thing is for sure, no access providerhad unique content that that could causea rapid growth in subscriber numbers.

Most Internet Service Providers (ISPs)now acknowledge that broadband accesshas rapidly become a commodity, likeelectricity, that enables services andcontent to users that are not owned in anyway by their own ISPs.

Services like online banking and p2p filesharing are the main reasons why broad-band access is such an success. It issomehow strange to realize, thatsuppliers of bank services do profit frombroadband and they are even in a largedegree responsible for its success, butthey are not responsible for investmentsor maintenance of the copper infrastruc-ture.

The way p2p filesharing lead to thesuccess of broadband Internet is also aninteresting case. Despite the fact ISPs inEurope are not banks and are not napster-like super nodes, banking and filesharinghave become very important issues forISPs in Europe.

The ever increasing demand for a saferinternet and the calls from music andmovie industry to do more to respecttheir rights, claims that are partly backedby politicians from all over Europe,causes the access providers to becomemore aware of the possible impact ofnew access technologies.

Back in 2000 some analysts alreadypredicted that fast internet and thenumber of users would change over thenext years. Some predicted that applica-tions like online gaming or digital photosharing sites could trigger rapid chances.But no one predicted that broadband usewould lead to world wide discussionsabout DRM, eSecurity and copy-protec-tion schemes.

by Rashid Niamat, NLIP (Dutch InternetService Provider Association)Editor: Heather Weaver, CCLRC, UK

EURO-LEGAL

News about legal information relating to Information Technology from Europeandirectives, and pan-European legal requirements and regulations.


IN BRIEF

79

ISTI-CNR — Paolo Cignoni is thefirst recipient of the Eurographics 'YoungResearcher' award in recognition of hisoutstanding contributions to the develop-ment and use of advanced multiresolutiontechniques in computer graphics and

scientific visual-ization.

The scope ofCignoni's inter-ests ranges fromscientific visual-ization to LODtechniques. Hehas publishedmore than 60

research papers in international refereedjournals and conferences, including sixpapers at Eurographics, and three papersat Siggraph'04. In addition, his work hasfound important applications in 3D scan-ning and cultural heritage.

Among his more recent results there is anovel approach to variable resolutionrendering based on a 'GPU-friendly' datastructure that is one or two orders ofmagnitude faster than previous solutionsand has been proposed for the efficientrendering of planet-sized digital terrainsand huge 3D models.

Paolo Cignoni is one of the foundingmembers of the Visual Computing Lab atISTI-CNR and a lecturer at the ComputerScience Dept. of the University of Pisa.

INRIA — Gilles Kahn, Chairmanof INRIA and INRIA's representative onERCIM's board of directors wasappointed member at the French NationalStrategic Advisory Board on InformationTechnologies on 27 September 2004. TheStrategic Advisory Board on InformationTechnologies (CSTI — ConseilStratégique des Technologies del'Information) is chaired by the FrenchPrime Minister and composed of leadingentrepreneurs from industry and R&D,selected for their own competence. CSTIis responsible for presenting recommen-dations to Government related to strategicorientations in innovation and researchand development, concerning informationtechnologies.

SpaRCIM — The First Spanish Conference on Informatics (CEDI 2005) isto take place between 13-16 September, 2005. While national conferences are tradi-tionally organized in several specific topics, in this case, all are to take place simulta-neously and in the same location. The main goal is to join the Spanish informaticsresearch community in order to demonstrate scientific advances, discuss specificproblems and improve the visibility of this area in Spain, while emphasizing its role inthe age of the information society.

CEDI will be organized in a 'federated conference' format, joining several morespecific symposia in areas including Artificial Intelligence, Software Engineering andDatabases, Programming Languages, Parallelism and Compute Architecture,Computer Graphics, Concurrency, Soft-Computing, Pattern Recognition, NaturalLanguage Processing, Ubiquitous Computing and Human-Computer Interaction.

In addition to the scientific activities of each symposium, the conference will organizea number of invited talks, round tables, social events and a gala dinner, at whichseveral awards will be announced, recognizing the efforts of some colleagues inpromoting the growth of the area in Spain.

More than 1000 researchers are expected to attend the conference. The event will befunded by the Spanish Ministry of Education and Science, and is sponsored bySpaRCIM.

The conference is organized by the Department of Computer Science at theUniversity of Granada. It will be chaired by Alberto Prieto from the University ofGranada, with Juan José Moreno Navarro (UPM and SpaRCIM director) as ProgramChair. Further details can be found at http://cedi2005.ugr.es/

CWI — CWI has started a new line of research on fundamental and practice-oriented cryptology. The institute is already known for the Number Field Sieve projectfor factoring large integers, which attracted world-wide attention with the factorizationof the RSA-512 internet security code. The new Cryptology and Information Securitygroup, led by prof. Ronald Cramer, includes this project. Additionally research is initi-ated in computational number theory with relevance for cryptology, public key cryptog-raphy, information theoretically secure cryptography, quantum cryptography, formalsecurity analysis, and applied information security. One of the group's major focalpoints is secure computation. This area deals with two or more parties who wish toachieve a joint task securely even though they are mutually distrustful and wish to keepsensitive, private information secret from each other. This is sometimes called multi-lateral security, as opposed to unilateral security in the case of secure communications.Practical applications include profile matching, joint database comparison, electronicvoting or auctions, and threshold cryptography. More information can be found onhttp://www.cwi.nl/pna5

FNR — The National Research Fund of Luxembourg launched the secondcall of its multiannual research programme 'Security and efficiency of new practicesin e-commerce for all socio-economic actors' (SECOM) on April 1st 2004. It has beenadopted by the Council of Government in the year 2000 and covers a total budget ofEUR 1.2 million over the period 2005 to 2007. Five project proposals were submitted bythe deadline of July 15th 2004 and are now being evaluated by independent experts. Theselected projects will start early 2005.

ERCIM – The European Research Consortium for Informatics and Mathematics is an organisation

dedicated to the advancement of European research and development, in information technology

and applied mathematics. Its national member institutions aim to foster collaborative work within

the European research community and to increase co-operation with European industry.

ERCIM is the European Host of the World Wide Web Consortium.

Fraunhofer ICT GroupFriedrichstr. 6010117 Berlin, GermanyTel: +49 30 726 15 66 0, Fax: ++49 30 726 15 66 19http://www.iuk.fraunhofer.de

Institut National de Recherche en Informatique et en AutomatiqueB.P. 105, F-78153 Le Chesnay, FranceTel: +33 1 3963 5511, Fax: +33 1 3963 5330http://www.inria.fr/

Swedish Institute of Computer ScienceBox 1263, SE-164 29 Kista, SwedenTel: +46 8 633 1500, Fax: +46 8 751 72 30http://www.sics.se/

Technical Research Centre of FinlandP.O. Box 1200FIN-02044 VTT, FinlandTel:+358 9 456 6041, Fax :+358 9 456 6027http://www.vtt.fi/tte

Swiss Association for Research in Information Technologyc/o Prof. Dr Alfred Strohmeier, EPFL-IC-LGL, CH-1015 Lausanne, SwitzerlandTel +41 21 693 4231, Fax +41 21 693 5079http://www.sarit.ch/

Slovak Research Consortium for Informatics andMathematics, Comenius University, Dept.of ComputerScience, Mlynska Dolina M, SK-84248 Bratislava, SlovakiaTel: +421 2 654 266 35, Fax: 421 2 654 270 41http://www.srcim.sk

Magyar Tudományos AkadémiaSzámítástechnikai és Automatizálási Kutató IntézetP.O. Box 63, H-1518 Budapest, HungaryTel: +36 1 279 6000, Fax: + 36 1 466 7503http://www.sztaki.hu/

Trinity CollegeDepartment of Computer Science,Dublin 2, IrelandTel: +353 1 608 1765, Fax: 353 1 677 2204http://www.cs.tcd.ie/ERCIM

Austrian Association for Research in ITc/o Österreichische Computer GesellschaftWollzeile 1-3, A-1010 Wien, AustriaTel: +43 1 512 02 35 0, Fax: +43 1 512 02 35 9http://www.aarit.at/

Norwegian University of Science and Technology Faculty of Information Technology, Mathematics andElectrical Engineering, N 7491 Trondheim, NorwayTel: +47 73 59 80 35, Fax: +47 73 59 36 28http://www.ntnu.no/

Spanish Research Consortium for Informaticsand MathematicsEsperanza Marcos, Rey Juan Carlos University, C/ Tulipan s/n, 28933-Móstoles, Madrid, Spain, Tel: +34 91 664 74 91, Fax: 34 91 664 74 90

Consiglio Nazionale delle Ricerche, ISTI-CNRArea della Ricerca CNR di Pisa, Via G. Moruzzi 1, 56124 Pisa, ItalyTel: +39 050 315 2878, Fax: +39 050 315 2810http://www.isti.cnr.it/

Centrum voor Wiskunde en InformaticaKruislaan 413, NL-1098 SJ Amsterdam, The NetherlandsTel: +31 20 592 9333, Fax: +31 20 592 4199http://www.cwi.nl/

Council for the Central Laboratory of the ResearchCouncils, Rutherford Appleton LaboratoryChilton, Didcot, Oxfordshire OX11 0QX, United KingdomTel: +44 1235 82 1900, Fax: +44 1235 44 5385http://www.cclrc.ac.uk/

Foundation for Research and Technology – HellasInstitute of Computer ScienceP.O. Box 1385, GR-71110 Heraklion, Crete, GreeceTel: +30 2810 39 16 00, Fax: +30 2810 39 16 01http://www.ics.forth.gr/FORTH

Czech Research Consortium for Informatics and MathematicsFI MU, Botanicka 68a, CZ-602 00 Brno, Czech RepublicTel: +420 2 688 4669, Fax: +420 2 688 4903http://www.utia.cas.cz/CRCIM/home.html

Fonds National de la Recherche6, rue Antoine de Saint-Exupéry, B.P. 1777L-1017 Luxembourg-KirchbergTel. +352 26 19 25-1, Fax +352 26 1925 35http:// www.fnr.lu

FWOEgmontstraat 5B-1000 Brussels, BelgiumTel: +32 2 512.9110http://www.fwo.be/

Order FormIf you wish to subscribe to ERCIM News

free of chargeor if you know of a colleague who would like to

receive regular copies of ERCIM News, please fill in this form and we

will add you/them to the mailing list.

Send, fax or email this form to:

ERCIM NEWS2004 route des Lucioles

BP 93F-06902 Sophia Antipolis Cedex

Fax: +33 4 9238 5011E-mail: [email protected]

Data from this form will be held on a computer database.

By giving your email address, you allow ERCIM to send you email

I wish to subscribe to the

❏ printed edtion ❏ online edition (email required)

Name:

Organisation/Company:

Address:

Post Code:

City:

Country

E-mail:

✄

You can also subscribe to ERCIM News and order back copies by filling out the form at the ERCIM website at http://www.ercim.org/publication/Ercim_News/

FNRSrue d'Egmont 5B-1000 Brussels, BelgiumTel: +32 2 504 92 11http://www.fnrs.be/

Grids: The Next Generation · 2008. 12. 8. · by Sarah Pearce, University of London, UK 18 The...

Documents

Transcript of Grids: The Next Generation · 2008. 12. 8. · by Sarah Pearce, University of London, UK 18 The...