High-level Composition of QoS-aware Grid Workflows: An Approach that Considers Location Affinity...

18
High-level Composition of QoS- aware Grid Workflows: An Approach that Considers Location Affinity Ivona Brandic, Sabri Pllana, Siegfried Benkner Institute of Scientific Computing (ISC) University of Vienna Austria The Workshop on Workflows in Support of Large-Scale Science (works06) In conjunction with HPDC 2006, June 20, Paris, France

Transcript of High-level Composition of QoS-aware Grid Workflows: An Approach that Considers Location Affinity...

Page 1: High-level Composition of QoS-aware Grid Workflows: An Approach that Considers Location Affinity Ivona Brandic, Sabri Pllana, Siegfried Benkner Institute.

High-level Composition of QoS-aware Grid Workflows: An Approach that

Considers Location Affinity

Ivona Brandic, Sabri Pllana, Siegfried Benkner

Institute of Scientific Computing (ISC)University of Vienna

Austria

The Workshop on Workflows in Support of Large-Scale Science (works06) In conjunction with HPDC 2006, June 20, Paris, France

Page 2: High-level Composition of QoS-aware Grid Workflows: An Approach that Considers Location Affinity Ivona Brandic, Sabri Pllana, Siegfried Benkner Institute.

Ivona Brandic works06, HPDC06

ProjectInvolvement

Quasi-real time neurosurgery support

Maxillo-facial surgery support

Radiosurgery: Neurosurgery support

Inhaled drug delivery simulation

Advanced Models, Applications and Software Systems for High Performance Computing

http://www.vcpc.univie.ac.at/aurora

http://www.gemss.de/

Page 3: High-level Composition of QoS-aware Grid Workflows: An Approach that Considers Location Affinity Ivona Brandic, Sabri Pllana, Siegfried Benkner Institute.

Ivona Brandic works06, HPDC06

Motivation

CT image reconstruction

Maxillo facial surgery simulation

QoS specification:time, price

maxTime = 5 hprice = 20 €

Quasi-real time neurosurgery support

Page 4: High-level Composition of QoS-aware Grid Workflows: An Approach that Considers Location Affinity Ivona Brandic, Sabri Pllana, Siegfried Benkner Institute.

Ivona Brandic works06, HPDC06

MotivationQoS specification:time, price

maxTime = 5 hprice = 20 €

Legal and securityaspects: „Location affinity“

CT image reconstruction

Quasi-real time neurosurgery support

Maxillo facial surgery simulation

Page 5: High-level Composition of QoS-aware Grid Workflows: An Approach that Considers Location Affinity Ivona Brandic, Sabri Pllana, Siegfried Benkner Institute.

Ivona Brandic works06, HPDC06

Outline

• BPEL extensions for the specification of „location affinity“ (QoWL)

• Use cases for „location affinity“• UML-based Domain Specific Language (DSL) for QoS-

aware Grid workflows• System architecture• Case study• Conclusion

Page 6: High-level Composition of QoS-aware Grid Workflows: An Approach that Considers Location Affinity Ivona Brandic, Sabri Pllana, Siegfried Benkner Institute.

Ivona Brandic works06, HPDC06

QoWL Code Example

<qowl-element name=„activityName“ portType=„...“ wsdl=„...“ ...>

...

<qos-constraints reqDescVar=„...“ ...>

<qos-constraint name=„beginTime“ value=„...“ weight=„...“ />

<qos-constraint name=„endTime“ value=„...“ weight=„...“ />

<qos-constraint name=„price“ value=„...“ weight=„...“ />

<qos-constraint name=„geographicAffinity“ value=„...“ />

<qos-constraint name=„gridSiteAffinity“ value=„...“ />

...

</qos-constraints>

</qowl-element>

any QoWL element any QoWL element

QoS extension of a BPEL elementQoS extension of a BPEL element

Page 7: High-level Composition of QoS-aware Grid Workflows: An Approach that Considers Location Affinity Ivona Brandic, Sabri Pllana, Siegfried Benkner Institute.

Ivona Brandic works06, HPDC06

Types of Location Affinity

• Grid site - gridSiteAffinity– Security, law, performance reasons

• Organisation - organisationAffinity– Hide business information from competitors– Sensitive business data

• Geographical region - geograficAffinity– Legal requirements – Electronic transfer of medical data – Medical studies including demographic data

Page 8: High-level Composition of QoS-aware Grid Workflows: An Approach that Considers Location Affinity Ivona Brandic, Sabri Pllana, Siegfried Benkner Institute.

Ivona Brandic works06, HPDC06

Specification of Location Affinity

A1A1

A2A2

A3A3

A4A4

A6A6

A5A5

A7A7

A8A8

A9A9

A10A10 A11A11

A12A12

A13A13

A14A14

A15A15

G1

G2

G3

<qowl-element name=„A3“ ...> ... <qos-constraints ...> <qos-constraint name=„gridSiteAffinity“ value=„SID“/> </qos-constraints></qowl-element>

<qowl-element name=„A11“ ...> ... <qos-constraints ...> <qos-constraint name=„geographicAffinity“ value=„GID“/> </qos-constraints></qowl-element>

<qowl-element name=„A7“ ...> ... <qos-constraints ...> <qos-constraint name=„organisationAffinity“ value=„OID“/> </qos-constraints></qowl-element>

Page 9: High-level Composition of QoS-aware Grid Workflows: An Approach that Considers Location Affinity Ivona Brandic, Sabri Pllana, Siegfried Benkner Institute.

Ivona Brandic works06, HPDC06

UML Modeling of QoS-aware Grid Workflows

• XML based workflow specification – error-prone and difficult for end users– Improvement by using a visual and intuitive modeling

language e.g. UML

• High level modeling based on UML– Activity diagrams– Visual modeling of QoS-aware Grid workflows– UML extensions mechanisms – Stereotypes and tagged values– UML based Domain Specific Language (DSL)

Page 10: High-level Composition of QoS-aware Grid Workflows: An Approach that Considers Location Affinity Ivona Brandic, Sabri Pllana, Siegfried Benkner Institute.

Ivona Brandic works06, HPDC06

Customizing UML for Modeling of QoS-aware Grid Workflows

<<metaclass>>Action

<<stereotype>>Invoke

portType: Stringwsdl: Stringoperation: StringinputVar: StringoutputVar: StringqosConstraints: QTypes

<<invoke>>Start

{ portType = appex wsdl = http://bridge:9355/SPECT/appex?wsdl operation = start inputVar = startRequest qosConstraints= qti:QTypes }

Definition of DSL e.g. Invoke Element

Usage of Invoke Element

UML based modeling of QoS-aware Grid workflows

A1A1 A2A2

A3A3 A4A4

„plain“ UML

StartStart

<<invoke>>Start

<<invoke>>Start

Page 11: High-level Composition of QoS-aware Grid Workflows: An Approach that Considers Location Affinity Ivona Brandic, Sabri Pllana, Siegfried Benkner Institute.

Stereotype Base Class Tags Description

Process «Process»

Activity qosConstraints:QType,variables:VType

Indicates that Activity represents a workflow process

Invoke«Invoke»

Action qosConstraints:QType,portType:String,operation:String,inputVar:String,outputVar:String,wsdl:String

Indicates that Action represents the operationInvocation of an external Grid Service

Copy«Copy»

Action From:FromTypeTo:ToType

Indicates that Action represents the data valueassignment

Sequence«Sequence»

SequenceNode qosConstraints:QType Indicates that SequenceNode represents a series ofactions which are executed sequentially

Flow«Flow»

StructuredActivityNode qosConstraints:QType Indicates that SequenceNode represents a set of Actions which may be executed concurrently

Receive«Receive»

AcceptEventAction portType:String,operation:String,variable:Stringwsdl:String

Indicates that AcceptEventAction represents a blocking message receive

Reply«Reply»

SendSignalAction portType:String,operation:String,variable:Stringwsdl:String

Indicates that SendSignalAction represents the replymessage to a message that was received through a«Receive»

Switch«Switch»

DecisionNode qosConstraints:QType Indicates that DecisionNode represents the conditionalexecution

While«While»

LoopNode condition:Boolean Indicates that LoopNode represents a while loop. Theloop body is executed until the condition is violated.

Page 12: High-level Composition of QoS-aware Grid Workflows: An Approach that Considers Location Affinity Ivona Brandic, Sabri Pllana, Siegfried Benkner Institute.

Ivona Brandic works06, HPDC06

QoWL Representation with the UML-based DSL (I)

A1A1 A2A2

<<Switch>>SampleSwitch

<<Switch>>SampleSwitch<<Switch>>

SampleSwitch

{qosConstraints = qtiQTypes}

Aggregation function <<Switch>>:

time = Max {time(Bi)| i=1,..,k}

price = Max {price(Bi)| i=1,..,k}

Switch

[case condition] [otherwise]

A1A1

<<Process>>SampleProcess

<<Process>>SampleProcess<<Process>>

SampleProcess

{qosConstraints = qtiQTypes}

Process

Page 13: High-level Composition of QoS-aware Grid Workflows: An Approach that Considers Location Affinity Ivona Brandic, Sabri Pllana, Siegfried Benkner Institute.

Ivona Brandic works06, HPDC06

A1A1

A2A2

A3A3

<<Sequence>>SampleSequence

<<Sequence>>SampleSequence<<Sequence>>

SampleSequence

{qosConstraints = qtiQTypes}

<<Sequence>> Aggregation function:

time = ∑ time(Ai)i=1

n

price = ∑ price(Ai)i=1

n

A1A1 A2A2

<<Flow>>SampleFlow

<<Flow>>SampleFlow<<Flow>>

SampleFlow

{qosConstraints = qtiQTypes}

<<Flow>> Aggregation function:

time = Max {time(Ai)| i=1,..,n}

price = ∑ price(Ai)i=1

n

SequenceFlow

QoWL Representation with the UML-based DSL (II)

Page 14: High-level Composition of QoS-aware Grid Workflows: An Approach that Considers Location Affinity Ivona Brandic, Sabri Pllana, Siegfried Benkner Institute.

Ivona Brandic works06, HPDC06

QoS-aware Grid Workflow System

QoS-aware Grid Workflow Engine (QWE)

XML parser / unparser

XML parser / unparser

QoSNegotiator

QoSNegotiator

QoSExecuter

QoSExecuter

WF Planner(static/dynamic)

WF Planner(static/dynamic)

Service Deployerand Generator

Service Deployerand Generator

Grid Infrastructure

TomcatTomcat Apache AXISApache AXIS

QoS-awareVGE Services

Service 1Service 1 Service nService n...

nonVGE Services

Service 1Service 1 Service nService n...

Teuta

GUIGUI

ModelTraverser

ModelTraverser

ModelCheckerModel

Checker

Page 15: High-level Composition of QoS-aware Grid Workflows: An Approach that Considers Location Affinity Ivona Brandic, Sabri Pllana, Siegfried Benkner Institute.

Ivona Brandic works06, HPDC06

Specification of MFSS Workflow

Page 16: High-level Composition of QoS-aware Grid Workflows: An Approach that Considers Location Affinity Ivona Brandic, Sabri Pllana, Siegfried Benkner Institute.

Ivona Brandic works06, HPDC06

Generated MFSS Code<sequence name="FEM_Sequence">... <copy name="CID"> <from part=“CID" variable=“getCIDRequest"/> <to part=“CID" variable=“uploadOperation"/> </copy> ... <invoke inputVar="CID" name="StartOperation" operation="start" portType="ApplicationExecutor"> <qos-constraints ReqDesc="maxilloReqDescVar"> <registry wsdl="http://gescher.univie.ac.at:9357/registry/reg?wsdl"/> <registry wsdl="http://kim.univie.ac.at:9357/registry/reg?wsdl"/> <registry wsdl="http://aurora.tuwien.ac.at:9357/registry/reg?wsdl"/> <qos-constraint name="beginTime" value="2006-02-02T16:00:00.000+02:00“ weight="0.3"/> <qos-constraint name="endTime" value="2006-02-02T18:00:00.000+02:00"/> <qos-constraint name="price" value="15" weight="0.7"/> <qos-constraint name="geographicAffinity" value="AT"/> </qos-constraints> </invoke> <invoke name="DownloadOperation“ ... /> ... </sequence>

Page 17: High-level Composition of QoS-aware Grid Workflows: An Approach that Considers Location Affinity Ivona Brandic, Sabri Pllana, Siegfried Benkner Institute.

Ivona Brandic works06, HPDC06

Related Work

• Languages– BPEL – GPEL– GSFL– SWFL– WSFL– (X)Scufl– AGWL– …

• Projects– BioOpera– Chimera– DAGMan– DiscoveryNet– FreeFluo– Grid ANT– Gridbus– JOpera– Kepler– LEAD– Pegaus– Taverna– Teuta– Triana– …

Page 18: High-level Composition of QoS-aware Grid Workflows: An Approach that Considers Location Affinity Ivona Brandic, Sabri Pllana, Siegfried Benkner Institute.

Ivona Brandic works06, HPDC06

Conclusion and Future Work

• Workflow execution should meet users requirements– Developed QoS extensions for BPEL– Developed the concept of „location affinity“– Performance, economical, security and legal QoS

• Simplification of workflow specification– Based on UML standard– Developed DSL for QoS-aware Grid workflows

• Validation through specification of a real world workflow• Future work

– Optimization and adaptivity issues