Network management survey - NIST...6."NetworkManagementSurvey" I.W.Cotton February,1974...

NBS TECHNICAL NOTE 805

Network Management Survey

NATIONAL BUREAU OF STANDARDS

The National Bureau of Standards 1 was established by an act of Congress March 3, 1901.

The Bureau's overall goal is to strengthen and advance the Nation's science and technology

and facilitate their effective application for public benefit. To this end, the Bureau conductsresearch and provides: (1) a basis for the Nation's physical measurement system, (2) scientific

and technological services for industry and government, (3) a technical basis for equity in trade,

and (4) technical services to promote public safety. The Bureau consists of the Institute for

Basic Standards, the Institute for Materials Research, the Institute for Applied Technology,

the Institute for Computer Sciences and Technology, and the Office for Information Programs.

THE INSTITUTE FOR BASIC STANDARDS provides the central basis within the UnitedStates of a complete and consistent system of physical measurement; coordinates that system

with* measurement systems of other nations; and furnishes essential services leading to accurate

and uniform physical measurements throughout the Nation's scientific community, industry,

and commerce. The Institute consists of a Center for Radiation Research, an Office of Meas-

urement Services and the following divisions:

Applied Mathematics — Electricity — Mechanics — Heat — Optical Physics — NuclearSciences 2 — Applied Radiation " — Quantum Electronics ' — Electromagnetics 3 — Timeand Frequency 3 — Laboratory Astrophysics 3 — Cryogenics 3 .

THE INSTITUTE FOR MATERIALS RESEARCH conducts materials research leading toimproved methods of measurement, standards, and data on the properties of well-characterized

materials needed by industry, commerce, educational institutions, and Government; provides

advisory and research services to other Government agencies; and develops, produces, anddistributes standard reference materials. The Institute consists of the Office of StandardReference Materials and the following divisions:

Analytical Chemistry — Polymers — Metallurgy — Inorganic Materials — ReactorRadiation — Physical Chemistry.

THE INSTITUTE FOR APPLIED TECHNOLOGY provides technical services to promotethe use of available technology and to facilitate technological innovation in industry and

Government; cooperates with public and private organizations leading to the development of

technological standards (including mandatory safety standards), codes and methods of test;

and provides technical advice and services to Government agencies upon request. The Instituteconsists of a Center for Building Technology and the following divisions and offices:

Engineering and Product Standards — Weights and Measures — Invention and Innova-tion — Product Evaluation Technology — Electronic Technology — Technical Analysis— Measurement Engineering — Structures, Materials, and Life Safety 4 — BuildingEnvironment * — Technical Evaluation and Application * — Fire Technology.

THE INSTITUTE FOR COMPUTER SCIENCES AND TECHNOLOGY conducts researchand provides technical services designed to aid Government agencies in improving cost effec-tiveness in the conduct of their programs through the selection, acquisition, and effective

utilization of automatic data processing equipment; and serves as the principal focus within

the executive branch for the development of Federal standards for automatic data processing

equipment, techniques, and computer languages. The Institute consists of the following

divisions:

Computer Services — Systems and Software — Computer Systems Engineering — Informa-tion Technology.

THE OFFICE FOR INFORMATION PROGRAMS promotes optimum dissemination andaccessibility of scientific information generated within NBS and other agencies of the FederalGovernment; promotes the development of the National Standard Reference Data System and

a system of information analysis centers dealing with the broader aspects of the National

Measurement System; provides appropriate services to ensure that the NBS staff has optimumaccessibility to the scientific information of the world. The Office consists of the followingorganizational units:

OffiV? of Standard Reference Data — Office of Information Activities — Office of TechnicalPublications — Library — Office of International Relations.1 Headquarters and Laboratories at Gaithersburg, Maryland, unless otherwise noted; mailing address

Washington, D.C. 20234.a Part of the Center for Radiation Research.3 Located at Boulder, Colorado 80302.* Part of the Center for Building Technology.


Ira W. Cotton

Computer Systems Engineering Division

Institute for Computer Sciences and TechnologyNational Bureau of StandardsWashington, D.C. 20234

National Byresti of Stan

APR 2 9 1974

Sponsored by the

National Science Foundation

18th and G Street, N.W.Washington, D.C. 20550

U.S. DEPARTMENT OF COMMERCE, Frederick B. Dent, Secretary

NATIONAL BUREAU OF STANDARDS, Richard W. Roberts, D/recfor

Issued February 1974

National Bureau of Standards Technical Note 805

Nat. Bur. Stand. (U.S.), Tech. Note 805, 91 pages (Feb. 1974)

CODEN: NBTNAE

U.S. GOVERNMENT PRINTING OFFICEWASHINGTON: 1974

For sale by the Superintendent of Documents, U.S. Government Printing Office, Washington, D.C. 20402

(Order by SD Catalog No. C13.46:805). Price $1.20

Foreword

This report Is one of a series of publicationsproduced by the Institute for Computer Sciences andTechnology, National Bureau of Standards, under GrantAG-350 from the National Science Foundation.

This grant supports a broad program of InvestigationInto the foundations of computer networking in support ofscientific and related educational efforts.

A listing of completed and planned publicationsproduced by the Institute under this grant follows*

1. "Primary Issues In User Needs"D. W. FifeChapter 10 in

MIT Press, Cambridge, MassachusettsExpected Publication* January 1 974

2. "Some Technical Considerations for ImprovedService to Computer Users"

T. N. Pyke, Jr.C0MPC0N 1973Seventh Annual IEEE Computer Society

International Conference

3. "Computer Networking Technology - AState of the Art Review"

T. N. Pyke, Jr. and R. P. BlancCOMPUTER MagazineComputer Society of the IEEEAugust, 1973

4. "Review of Network Management Problems and Issues"A. J. NeumannOctober, 1973NBS Technical Note No. 795

5. "Annotated Bibliography of the Literature onResource Sharing Computer Networks"

R. P. Blanc, I. W. Cotton, T. N. Pyke, Jr.zr.4 S. W. Wat kins

September, 1973NBS Special Publication 38k

in

6. "Network Management Survey"I. W. CottonFebruary, 1974NBS Technical Note No. 805

7. "User Procedures Standardization forNetwork Access"

A. J, NeumannOctober, 1973NBS Technical Note No. 799

8. "Review of Computer Networking Technology"R. P. BlancJanuary, 1974NBS Technical Note No. 804

9. "Microeconomics and the Market forComputer Services"

I. W. CottonSubmitted for publication

10. "Cost Analyses for Computer Communications"R. P. BlancNBS Report In Preparation

11. "Network User Information Support"A. J. NeumannNBS Technical Note No. 802Expected Publications February 1974

12. "Quality Service Assurance Experiments"R. Sti 1 lmanJanuary, 1974NBS Technical Note No. 800

13. "A Guide to Networking Terminology"A. J. NeumannNBS Technical Note No. 803Expected Publication* February 1974

14. "Research Considerations in Computer Networking'D. W. FifeNBS Technical Note No. 801Expected Publication* February 1974

IV

CONTENTS

Page

Introduction 1

ARPA Network k

MERIT Network 22

Triangle Universities Computation Center 34

Oregon State Regional Network 51

TYMNET 61

Conclusions 71

Acknowl edgements

The contributions to this document of Robert Blanc,Albrecht Neumann, Dennis Fife and Thomas Pyke, Jr., allof the Institute for Computer Sciences and Technology,are expressly and gratefully acknowledged. Also, theassistance of technical and managerial personnel fromeach of the networks surveyed, both in providinginformation and in reviewing drafts, is noted andappreciated. Of course, the ultimate responsibilityfor the content of this document is the author's alone.

VI


Ira W. Cotton

This report presents the results of a study of managementpractices In different computer networks. Five networkswere chosen as typical of different approaches to networkImplementation and management* Defense Advanced ResearchProjects Agency (ARPA) Network, MERIT Network, TriangleUniversities Computation Center (TUCC), Oregon StateRegional Network and Tymnet, a commercial network. Acommon format Is employed to survey each network. Whilethe report Is not Intended to be prescriptive, someempirical observations are presented for each topiccovered.

Key words * Computer network; management;network; network management

Introduction

This report on management practices In different computernetworks presents some of the results of a study conductedby the Institute for Computer Sciences and Technology,National Bureau of Standards, for the National ScienceFoundation In support of the "Networking for Science"program. The approach to management taken in this reportIs positive (i.e., reporting on current practices as theyexist); the normative or prescriptive approach is taken Ina separate report as indicated in the foreword.

This survey Is meant to be representative and illustrativerather than exhaustive or conclusive. The networkscovered were chosen to be typical of different approachesto network Implementation and management.

The ARPA network Is a large distributed network ofautonomous, heterogeneous computer systems. It hasfocused on the development of network technology andresource-sharing techniques, rather than organizationalprobl ems.

Network Management Survey Introduction

MERIT is a controlled experiment in networking on aregional basis with heterogeneous computer systems.Considerable attention has been focused during thenetwork's development on the organizational problems.

TUCC Is another regional network, but with homogeneouscomputers and a larger number of participatinginstitutions. TUCC has also given considerableattention to organizational issues.

The Oregon State network is representative of manycentralized or "star" networks serving a regionalcl i entele.

TYMNET is the only commercial network included in thestudy. It was desired to include a profit-seekingnetwork in the study, so as to be able to comparemanagerial practices in non-competitive environmentswith actual business practices. (Of course, theinclusion of TYMNET in this study in no way impliesendorsement of this network).

Information for this report was gathered from primary andsecondary literature sources, from telephone and on-siteinterviews, and by iterative review of thi s -document byrepresentatives of the networks covered. In some casesthe information presented here will already be out of dateby the time this report is published, so rapidly are thesenetworks developing.

A common format was employed for presenting the details ofeach network. The table of contents which follows is anoutline for each of the five surveys which follow it.Network architecture is separated from network management,and the latter is broken down into a number of differentfunctional areas. This approach was adopted to permiteasy comparison of specific managerial concerns fromnetwork to network. Some empirical observations from suchcomparisons are presented in a concluding section whichfollows the same format as each of the reports.

Some comments are probably in order regarding the use ofthe term "successful" with reference to the networkssurveyed. In the author's view, success is measured bythe satisfaction and acceptance of the network and itsfacilities by its users. This should translate to thecontinued existence of the network, and in most cases togrowth. Used in this sense, the term transcends meretechnical accomplishments. A network which performs

Network Management Survey Introduction

adequately in the technical sense but which fails tosatisfy user needs can not be considered successful.

Table 1. Organization of Each Survey

I. Introduction

II. Network Identification and General DescriptionA. Sponsoring OrganizationB. Purpose/mission of NetworkC. Status and TopologyD. Technology SummaryE. Future Plans

III. Network OrganizationA. Structure and ExtentB. Functions Performed

1 . PI anni ng2., Instal 1 at i on3. OperationsA. User Services

C. Interfaces ( rel at i onshi ps with other organizations)

IV. Financial and Legal ConcernsA. Capi tal i zat i onB. AccountingC. RegulationD. Security

V. ConclusionsA. Summary of ProblemsB. Lessons Learned

AnnexesA. Bi bl i ographyB. Special Vocabulary

ARPA Network

I, Introduction

The ARPA Network is characteristic of largedecentralized networks of autonomous computer systems.Its continued successful operation has been widelypublicized in the professional literature and tradepress, and has demonstrated the technical feasibilityof packet-switching for large networks.

II. Network Identification and General Description

A. Sponsoring Organization

The network is sponsored by the Defense AdvancedResearch Projects Agency (ARPA), an agency of theDepartment of Defense concerned with research anddevelopment in areas of advanced technology.

B. Purpose/mission of Network

The ARPA Network was begun as a research effort toinvestigate multiple computer resource sharing andto demonstrate the feasibility of packet switchingtechnology.

The network currently supports many ARPA-sponsoredresearch programs by providing access to resourcesnot available locally.

C. Status and Topology

The network has ,een operational since mid-1971.Figures 1 and 2 illustrate respectively the recentgeography and topology of the network . Thenetwork currently includes over 30 sites with over40 independent computer systems connected.

Reliability of the various network componentsvaries widely. Currently the communications subnetof Interface Message Processors (IMPs) and circuitshas extremely high reliability, with onlyoccasional outages for particular IMPs or circuits.Since most nodes are connected to at least twoother nodes, the outage of a single circuit doesnot disturb network operation. The service sitesare generally less reliable than the subnet,although the statistics vary widely from site tosi te.

0)

ccUJCD

IDI-0.UJin

or<

oiQ.<CC

OoUJ

o*rccoB:i-LU

<Q-

CC<

DuH

CO

a>

ccLUCO

LUI-Q.IDC/)

CL<

<Ooo

cco

uuz

2<

CM

DOH

Network Management Survey ARPA Network

D. Technology Summary

The ARPA Network may be characterized as adistributed store-and-forward network ofheterogeneous computer systems (hosts). Hosts areconnected to the communications subnet by means ofa software interface called the Network ControlProgram (NCP) and a hardware interface which mayhave the characteristics of either a channel or acommunications line. Each host is connected to aswitching center in the subnet called an InterfaceMessage Processor (IMP) which contains an augmentedHoneywell 516 or 316 computer; up to fourindependent hosts may be connected to the same IMP.Each IMP is connected to two or more neighboringIMPs by means of dedicated 50kb communicationslines. Host to host messages are passed from thesending host to its IMP, where they are broken intopackets and relayed to their destination by thesubnetwork of IMPs and communications lines. Therouting is adaptive; i.e., the route of any givenpacket is not established in advance and in generalthe several packets of a message may followdifferent routes. The destination IMP willreassemble the message and deliver it to the properhost.

Protocols in the network are constructed accordingto a layered approach. The lowest level protocolis a binary synchronous communications protocolgoverning traffic exchange between IMPs. Theso-called "first level" protocol governs thelogical exchange of information between host andIMP. The "second level" protocol governs thelogical exchange of information between NetworkControl Programs in communicating hosts. The"third level" of protocol refers to anycommunications occur Ing between processes in thehost machines. Such third level protocols includethe Initial Connection Protocol (ICP), datatransfer protocol, file transfer protocol, remotejob entry protocol, graphics protocol, and others.

A special third level protocol called TELNETdefines a network virtual terminal and permits allterminals on the network to provide a similarinterface to processes in a separate host computersystem. A special IMP which is augmented by theaddition of memory and a multiline controller (a


specially designed component containing centrallogic and line interface units) can provide directnetwork access to terminals without going through aseparate host computer system by providing theTELNET function itself. Such an IMP is called aTerminal IMP, or TIP.

E. Future Plans

The ARPA network continues to grow at a rate ofabout one site per month. It has been extendedinto Hawaii and Europe via satellite links.Network traffic has been growing at an exponentialrate for some time. The main technical workincludes redesign of the present IMPs, developmentof mi ni -computer front ends to IMPs for remote jobentry and private line interfaces, and thecontinued refinement of the protocols.Organizationally, the network is turning itsattention to improving resource-sharing facilities.

III. Network Organization

A. Structure and Extent

Figure 3 illustrates the present organizationalstructure for the network.

The Advanced Research Projects Agency (ARPA) of theDepartment of Defense initially conceived andfunded the network, and presently directs itsoperation through the Information ProcessingTechniques Directorate. ARPA is not a largeagency, serving primarily as a granting agency, andonly several people spend their full time onnetwork activities.

ARPA has turned over the day-to-day operation ofthe communications network to the USAF RangeMeasurements Laboratory (RML) at Patrick AFB inFlorida. RML will serve as the procurement agentfor IMPs and maintenance services and as thecentral liaison point for the Defense CommercialCommunications Office and the common carriers (seebelow, III-B-1), Bolt Beranek and Newman, NetworkAnalysis Corporation, and prospective new users.

Bolt Beranek and Newman (BBN) of Cambridge,Massachusetts, is the primary ARPA contractor in

8

HONEYWELL n

RRM

~ RESEARCH SITESDDR

- SERVICE SITES

NCC

ATT—

mm

1

L USER SITES

MAC— ARPA

|

NMC— 1

RML— 'NETWORK FACILITATORS"

NIC— HML / MITRE CORPORATION

l

PROSPECTIVE USERS

BBN- Bolt Beranek & Newman

ATT- American Telephone & Telegraph

NAC- Network Analysis Corporation

NMC- Network Measurement Center (at UCLA)

NIC- Network Information Center (at SRI)

NCC- Network Control Center (at BBN)

RML- Range Measurements Laboratory (Patrick AFB)

Figure 3. Present Organizational Stricture for Operation of the ARPA Network


the development and operation of the network. BBNdesigned and constructed the IMPs and TIPs whichcomprise the subnet, wrote the software for theseprocessors, and participated in the specificationof the network protocols. BBN currently overseesall network modifications, dealing, through RML,with the Bell System for the wideband circuits, andwith Honeywell Information Systems for theprocurement of H-516 and H-316 processors forinclusion in the IMPs and TIPs. BBN also operatesa Network Control Center (NCC) which monitors theoperation of the network on a round-the-clock basisand which aids in the diagnosis of failures andInitiates and coordinates maintenance efforts.

Honeywell Information Systems is an OEM supplier ofbasic H-316 and H-516 processors, from which BBNfabricates the IMPs and TIPs. Honeywell fieldengineers also assist in the installation of theIMPs and TIPs, and are responsible for maintenance.

Network Analysis Corporation (NAC), Glen Cove,N.Y., serves as an ARPA contractor for analyticalstudies to determine the optimum topology of thenetwork and select the topology actually used.

The Network Measurement Center (NMC) is anARPA-supported research group at the University ofCalifornia (Los Angeles) which performsmeasurements and investigates the performance ofthe network.

The Network Information Center (NIC) at StanfordResearch Institute provides a reference centerserving to receive, record, index, and transmit,online and offline, Information produced by andabout the Network. To do this it designsInformation-handling tools for dialogue andrecord-handling. The NIC currently operates aPDP-10 TENEX system which is used both fordevelopment and network services. However, it Isplanned to separate the development activities fromthe service functions through an arrangement withTymshare, Inc. The involvement with Tymshare isstrictly on a "facilities management" basis. Thatis, Tymshare will operate a computer systemvirtually identical to the one at the NIC, toprovide reliable Tenex, NLS, and Journal service toNIC users and other selected customers. The NIC

10


staff will remain at SRI, as will theresponsibility for further developments in theseareas

.

Approximately 18 hosts may be classified asresearch sites. These sites perform research In anumber of areas under ARPA contracts and utilizethe network in support of their research.

Six sites are classified as service sites: UCLA(360/91), SRI (NIC), UCSB (360/75), B8N (PDP-10),USC-ISI (PDP-10) and MIT (MULTICS). Some of thesesites offer services to the network community on afee basis; others are subsidized by ARPA.

The remaining sites are users, although intersitearrangements are often made for the sharing ofresources. Almost all of the sites make some oftheir resources available, but not all guarantee tointeract as a service organization. Each site hasa Technical Liaison to answer technical questions,and a Station Agent to handle questions concerningadministrative, documentation, and similarquestions regarding the network resources and theNIC.

The "network facilitators" were an informalcommitee of experienced users who proselytized andattempted to solve network problems for users.This group was In existence for about a year, afterwhich it disappeared. The performance of this typeof function has since been delegated to a group atthe Mitre Corporation under contract to ARPA.Mitre publishes a network newsletter, responds touser inquiries and provides various kinds ofassi stance.

Other groups that are in a sense "networkfacilitators" are these special and generalinterest groups »

ANTS Support GroupARPANET Satellite SystemComputer Based Instruction GroupFile Transfer Protocol Interest GroupImlac Interest GroupInternational Network Working GroupInternational Packet Network Working GroupNetwork Associates Group

11


Network Graphics GroupNetwork Information CenterNetwork Liaison GroupNetwork Measurement GroupNetwork Station Agent GroupNetwork Working Group Steering CommitteePrincipal InvestigatorsPacket Radio GroupRemote Job EntrySpeech Understanding Research GroupTIP Users GroupTNLS User GroupTENEX Users Group

B. Functions Performed

1 • PI anning

Network growth Is controlled by the sponsoringagency, ARPA. However, Karp (1973) indicatesthat the first 15 sites were ARPA researchfacilities selected for their expertise insoftware and system design rather than for thecomputer resources they could provide.Subsequent sites have provided both additionalresources and users of these resources. Theredoes not appear to be any fixed or publishedpolicy for determining what sites are to beadded. Network growth Is presently limited bythe rate at which Honeywell and BBN can supplyIMPs (about 1 per month).

When a new site is selected, Network AnalysisCorporation determines for ARPA the new topologyfor the subnet. The network topology is notoptimized each time a node is added, since thatmight entail too costly and disruptivealterations of existing circuits, but the newnode is added in as cost-effective a means aspossible. Occasionally, more extensive changesare made to the network topology as needswarrant. The planning function is thuscentralized and supported by appropriateanalysi s.

All hardware procurement for the network Iscontrolled by ARPA; the individual sites are notinvolved. ARPA places orders (through RML) forIMPs and TIPs with BBN. Communications circuits

12


are ordered through the Defense CommercialCommunications Office (DECCO) which deals withthe various common carriers. As explainedbelow, in practice BBN coordinates circuitinstallation with the carriers.

2. Instal 1 at i on

Each site which is added to the network isresponsible for fabricating the interface to theIMP or TIP and for coding a network controlprogram (NCP). Specifications for each areavailable from Bolt Beranek and Newman. Thehost organization is also responsible for sitepreparation (floor space, power, airconditioning, etc.) for the IMP, which willreside at the host's own location.

Bolt Beranek and Newman coordinates IMPinstallation through the Network Control Center.The normal installation team consists of a BBNrepresentative, the person from the locaiHoneywell office who will maintain the machine,possibly an additional person from the mainHoneywell office, and telephone companypersonnel

.

3. Operations

The local host organization is responsible formaintenance of its host processor, NCP and IMPinterface. The local Honeywell office willmaintain the IMP itself. AT&T Long LinesDivision is responsible for maintenance of themodems and communications circuits.

For diagnostic and control purposes, BBNoperates a special host system in Cambridge,Massachusetts, which is called the NetworkControl Center. The NCC regularly receivesstatus reports from all the IMPs in the networkregarding the operational status of theircommunication circuits and their neighbouringIMPs. Special programmable debugging and faultisolation procedures (such as looping lines backinto the same IMP) may be initiated remotelyfrom the NCC. Fault isolation may or may notrequire the assistance of local host personnel.

13


The maintenance function, which may involveseveral organizations including the Bell System,Honeywell, BBN and the local host, appearscomplicated, but is handled ef

f

icient ly *t. TheNetwork Control Center is very effective indiagnosing failures and coordinating maintenanceamong the various groups involved.

The division of responsibility for software issimilar to that for hardware. All IMP softwareIs controlled by BBN and Is currently loaded viathe network itself. Host organizations areforbidden to modify the software In their IMPs.

Functional protocols are specified by committeesof host, ARPA and contractor personnel. At thelower levels these are fairly well agreed uponand debugged, and hosts agree to abide by them.At the higher levels, the protocols are stillevolving, and subsets of hosts often experimentwith variants among themselves.

At the present time, all of the lower levelprotocols are sufficiently well defined anddebugged as to be of little or no concern tc theaverage user. The higher level protocols, ormore precisely, the lack of generally acceptedand debugged higher level protocols (forexample, remote job entry) have and continue tobe a major hindrance to increased networkutilization. However, these problems arerecognized and new and improved protocols arebeing constructed and tested by variouscontractors.

4. User Services

Users at one site seeking to utilize resourcesat another site are required to be familiar withthe characteristics (log-on procedures,operating system commands, program conventions,etc.) of that site. Documentation is, ?ngeneral, provided by the serving site accordingto its own conventions. Several organizationsdo exist, however, to help users access remoteresources.

The Network Facilitators Group was an informalcommittee of experienced personnel at various

14


sites around the country who organized topromote network utilization. This informalgroup has since been replaced by a group at theMitre Corporation who have an ARPA contract toassist network users.

The Network Information Center is operated byStanford Research Institute to facilitate thecollection and dissemination of data produced byand about the Network. The NIC maintains onlinefiles of data about the Sites and people on theNetwork, maintains online tools for access tothe data, and produces offline notebooks,indexes and directories of the data for use bythe Network and other networking agences. TheNIC also functions to reproduce, catalog, indexand distribute online and hardcopy documents asrequested by Network Sites in the process ofbuilding and using the Network.

The user support function Is probably the leastwel 1-provi ded function in the ARPA network. Thequality of documentation and the generalavailability of assistance from sites variestremendously from site to site. The networkfacilitators helped at selected user sites, butbecause of the Informality of their organizationand because they were not located at all sitesthey were not a general solution. Mitre hasbeen at work on their user assistance task fortoo brief a period to be evaluated. The NIC hasnot provided a general solution to thedocumentation problem, since they do not attemptto distribute all the documentation needed torun at a particular site nor to enforce anydocumentation standards.

C. Interfaces (relationships with other organizations)

The question of the ARPA network's future statushas yet to be settled. The network is a closedcommunity, available only to governmental agenciesand their contractors, but the network has beenconnected for demonstration purposes to other(commercial) networks (e.g., TYMNET). The ARPAnetwork is presently functioning as a marketplacefor the sale of computer time to the networkcommuni ty.

15


ARPA has indicated on several occasions that itdoes not intend to operate the network indefinitelyand has already turned over daily operation toanother governmental organization (RML). Variousoptions for converting the network to commercialoperation have been considered, but none have beenconsidered feasible. It remains to be seen whatthe future of the network will be.

In another vein, the technology developed for thenetwork has already received commercial attention.Several companies have filed applications with theFederal Communications Commission to set up andoperate similar networks as common carriers (insome cases, using basic communications circuitsprovided by other common carriers in a so-called"value-added" configuration).

Finally, ARPA does seem interested In investigatingthe means and the difficulties of connecting toother networks. Many members of the ARPANETcommunity participate in the International NetworkWorking Group, originally an ARPA-sponsoredinitiative which is now a working group under IFIPwith international participation. The purpose ofthe group is to develop inter-network protocols andto mount some international experiments.

IV. Financial and Legal Concerns

A. Capitalization

The network was funded by ARPA, so capitalizationIs provided through research grants. ARPAcontinues to subsidize the subnet communicationscircuits. Each of the IMPs and TIPs was eitherpaid for by ARPA or the participating hostorganization. In general, some of the early siteswere ARPA contractors, and so had their IMPsprovided, while many of the newer sites are userswho are paying for their own.

B. Accounting

Participation In the network requires access to thecommunications subnet and access to individualhosts on the network. IMPs and TIPs are availablefor a fixed fee from ARPA, which obtains them fromBBN. Monthly maintenance charges must be paid

16


after the first year of operation. Communicationscharges are assessed by ARPA according to usages abase fee plus a variable fee for traffic above agiven minimum. In fact, most network participantshave been supported by ARPA in one way or another,and communications costs continue to be subsidized.

Accounting for resources at the host sites is doneby each of the hosts concerned. Each user whodesires to utilize remote resources must open anaccount at the appropriate site. The friendlinessof sites to external users varies considerably, asdo billing rates. Billing procedures vary widely(some sites have been known to close an accountwhen it ran out of funds without notifying theuser) and require using sites to deal with manyvendors. The overall situation is recognized to beless than satisfactory, and a committee ofprincipal Investigators is currently trying todevelop a network-wide accounting scheme.

C. Regulation

No commercial users or non-research commercialservers are currently permitted on the network. Asa research activity sponsored by the Department ofDefense, the network is not subject to regulationby other government agencies (e.g., the FCC).

The wideband communications circuits used in thesubnet are leased by ARPA through the DefenseDepartment at bulk (TELPAK D) rates. These ratesrepresent considerable economies of scale for largedata communications users.

D. Security

The communications subnet will insure that messagesare delivered to the proper host. Thenon-deterministic routing of the individual packetsof a message could be viewed as providing somedegree of security to the subnet. At that point Itis the responsibility of the host to insuredelivery to the proper user. There is no checkingon the origin of messages to insure that they agreewith the identification given in a log-on sequence.

File security is the responsibility of theIndividual hosts. Log-on procedures, keyword

17


access and the like are among the proceduresemployed by the various hosts as protectivemechanisms.

V. Conclusions

A, Summary of Problems

Start-up Requirements - Fabrication of the IMPinterface and coding of the Network Control Programhave been major obstacles to new hosts joining thenetwork. BBN has fabricated some interfaces, butprefers not to do so. This problem is expected tocontinue, as the character of sites joining thenetwork changes from researcher to user. However,a mitigating factor may be the inventory ofinterfaces and NCPs which have already beenconstructed.

Cost - The question of cost is growing inimportance as the network operation is examined bypotential commercial suppliers and as the fullcosts become known. Published documentation oncosts have thus far dealt mainly with thecommunications subnet. Even here, the costs may bedistorted because published figures are derivedfrom loading factors which have never been realized(although traffic has been growing at anexponential rate). Also, the cost of networkaccess (an IMP or TIP) remains high (though lowercost replacements are currently under development).However, the most significant concern is that thecost of the subnet may not be the major networkcost. Recent studies have shown that the overheadassociated with the NCP may be substantially largerthan was previously believed. Additional coststudies are indicated, as well as a reevaluation ofthe current protocol strategy with a view toreducing overhead.

Reliability - In contrast to the now highreliability of the subnet, the host sites, at whichthe real work Is done, vary widely in reliability.

Heterogeneity - While some facilities (such asTELNET, network mail and file transfer protocol)have been provided to permit common operation onhosts of different types, the problem has not beensolved in general. Executive level commands are

18


all different for different machines, text editorsare different, log-on procedures are different,etc. This is a difficulty which networks have onlyexacerbated by making additional systems availableto potential users.

User Services - The need for readable, accurate,complete and available documentation cannot bestressed too much. Of equal importance, however,Is the occasional need for hand-holding. On-linetutorials may provide some relief, but personalassistance by knowledgable and friendly personnelwill never be completely replaced by documentation.The current level of these services on the ARPAnetwork leaves room for improvement.

Protocols - Lack of particular protocols such asgraphics and remote job entry has cut off potentialusage in some cases. This problem will be resolvedas these protocols become better developed and aremade generally available.

B. Lessons Learned

In fairness, it must be recognized that the ARPAnetwork began as an experiment in networking amongresearch-oriented sites. It has achieved itsobjective of demonstrating the feasibility of thepacket-switching approach. Large numbers of usersdepend on the network now for their day-to-daywork. Many of the problem^ which have beenidentified have arisen as the network matured andthe complexion of its participants changed fromresearch-oriented to usage-oriented.

In general, the network has functioned best wherethere has been formal responsibility andorganization, for example, at the Network ControlCenter. The less directed efforts have beencorrespondingly less successful, for example, thehigher level protocol committees.

The network seems also to have been most useful Inoffering services when the differences between hostsystems have been bridged automat icaly by the network.For example, perhaps the most successful service hasbeen the TELNET, which has permitted a wide range ofterminals to operate more or less identically on manydifferent host systems. The resources for which this

19


type of standardization has not been performed havebeen less successful In attracting wide utilization.Additional work in this area seems indicated.

Annexes

A. Bibliography

Carr, S., Crocker, S. and V. Cerf. "Host/hostcomunl cation protocol in the ARPA network." SJCC,1970, pp. 589-597.

Crocker, Stephen D. et. al . "Function-orientedprotocols for the ARPA computer network." SJCC1972, pp. 271-279.

Frank, Howard, Robert E. Kahn and LeonardKleinrock. "Computer communications network design— experience with theory and practice." SJCC1972, pp. 255-270.

Heart, Frank E. et. al. "The interface messageprocessor for the ARPA computer network." SJCC1970, pp. 551-567.

Kahn, Robert E. and William R. Crowther. "A studyof the ARPA network design and performance." BoltBeranek and Newman, Report No. 2161, August 1971.

Karp, Peggy M. "Origin, development and currentstatus of the ARPA network." , pp.49-52.

McKenzIe, Alexander A. et. al. "The networkcontrol center for the ARPA network." (Proc.ICCC), 1972, pp. 185-191.

MImno, N. W., et. al . "Terminal access to the ARPAnetwork^ experience and improvements.", pp. 39-43.

Network Analysis Corporation. "ARPANET* design,operation, management and performance." Report toRome Air Development Center, April 1973.

Ornstein, S. M. et. al . "The terminal IMP for theARPA computer network." SJCC 1972, pp. 243-254.

20


Roberts, Lawrence G. and Barry 0. Wessler."Computer network development to achieve resourcesharing." SJCC 1970, pp. 543-549.

B. Special Vocabulary

IMP - Interface Message Processor - A speciallymodified Honeywell 316 or 516 processor whichserves as the communications computer in thenetwork.

TIP - Terminal IMP - An IMP which is augmented byadditional memory and a multiline controller. TheTIP contains a network control program and a TELNETprogram within it to permit terminals to access thenetwork directly through it.

Multiline Controller - A specially designedmultiplexor-like device which supports the accessof up to 64 terminals of varying type into a TIP.

Message - A logical unit of data exchange betweenprocesses.

Packet - Physical segments of a message which arethe transmission units In the subnet.

Subnet - The array of IMPs, TIPs and communicationcircuits which deliver messages from one host toanother.

21

MERIT Network

I. Introduction

The MERIT Network Is representative of a controlledexperiment In networking on a regional basis. Thestraightforward design of the communi cat ions systemand the limited size of the network have permittedattention to be focused on the managerial problems ofthe network.



In 1966, Michigan State University, Wayne StateUniversity and the University of Michigan formallyestablished a program of mutual cooperation knownas MICIS (Michigan Interuni versi ty Comi ttee onInformation Systems). MICIS established anon-profit corporation, MERIT (Michigan EducationalResearch Triad, Inc.), for the purpose of receivingand distributing funds for research. It is thisnon-profit corporation which sponsors the network.


The Merit Network is a prototype educationalcomputing network that seeks to enhance theeducational and computing resources of eachuniversity by permitting network participants toshare resources. The objectives of the networkwere broadly stated as gaining, through thedevelopment and successful implementation of anetwork, knowledge about and solutions to theproblems of network operation in an establishededucational computing environment.

The three computer systems in the network aresufficiently different as to make desireable thedirecting of particular types of work to one ofthem from the others.

The University of Michigan's system, usingduplex IBM 360/67 hardware, was designedespecially for timesharing using MTS.

The Michigan State University computer (CDC6500) is unusually fast and thereforewell -suited to large, compute-bound jobs.

22

Network Management Survey MERIT Network

The Wayne State University Computer Center,using duplex IBM 370/67 hardware running the MTStimesharing system, has developed a specialcompetence in administrative data processing.

At the time of organization of the network, acooperative policy in acquiring special peripheralequipment was considered feasible. It wassuggested that relatively unusual equipment, suchas a film recorder-scanner, might be purchased byone installation to serve all three universities,


The network became operational in the summer of1972. The topology of the network is three node,fully connected. Presently the network providescommunication service, on a nearly continuous basiswhenever the host systems are up; usage through1973 has been light, however. Aupperle (1973)reports that current network-use data indicatesthat between one and two million bytes aretransmitted monthly by an average of twentyindividual users. About 100 different users havetried the network during the second half of 1972,the first six-month period of statistical datagather ing.


The three host systems in the MERIT network aretied together through small communicationscomputers located at each host site, which arethemselves interconnected by means of modemsthrough the switched telephone network. Theinterface which the communications computerspresent to the host system are uniquely adapted tothe requirements of each particular host; theinterface which the communications computerspresent to the telephone network (and thus to eachother) are identical. A somewhat novel capabilityIs the ability to dynamically vary the bandwidth ofthe communications paths available between pairs ofhosts. This is accomplished by providing eachcommunications computer with four separate modemsand an automatic calling unit. The bandwidthrepresented by the four lines may be allocated tocommunications with either of the other two hostsas dictated by Immediate communications

23


requirements; normally at least one line Is keptopen to each other remote site.

The communications computer consists of a standardDigital Equipment Corporation PDP-11/20 with 16Kwords of 16-bit core memory. Four different typesof interfaces are required:

1. Data set interface

2. Automatic calling unit and multiplexori nterf ace

3. IBM 360/67 interface (one for each of twohost systems)

4. CDC 6500 interface (for one host system)

The data set interfaces are designed to transferentire messages over the telephone network directlyto and from the PDP-11's core memory withoutprogram intervention, once the transfer isInitialized by software action. These Interfacesare designed to function as either half-duplex orfull-duplex units over a wide range of frequenciesand are compatible with binary synchronouscommunications procedures.

Automatic calling unit and multiplexor interfacesprovide the communications computer with theability to dial, under program control, a singletelephone call. The automatic calling unit may beshared among eight telephone lines.

The host Interfaces (IBM 360 and CDC 6500), likethe data set interfaces, are designed to transferentire messages without intervention, onceappropriate action is initiated. The interfacestransfer data in parallel between the host and thecommunications computer, and they partially resolvethe word length mismatch that exist between thesemachines. The host interfaces also cause thecommunications computer to appear as a number ofidentical, but separate, devices to the host, thussimplifying the logical structure of multiple useractivity over the network and minimizing the amountof the host's network support software.

The systems programming requirements for the MERIT

24


computer network consist of device suport for thecommunications computer in each of the hosts, andan operating system in the communications computerwhich provides support for the message switchingfunction and an interface to each host. Each hosttreats the communications computer as an I/Odevice. Together, the software elements in thehost and communications computers permit a user toestablish a path to a remote host and to utilizeservices there. All terminal access into thenetwork is through the local host; there is nodirect terminal access Into the network whichbypasses the local host. However, this capabilityIs being considered for future Implementation.

E. Future PI ans

The MERIT network continues to serve the threeparticipating universities. There is considerationfor a remote version of the communications computerto allow terminals to access the network directly(without there being a local host) so as to be ableto offer service to other institutions In theregion. Work is underway to extend thecommunications computer software to allow therouting of messages in a 1 ess-than-ful 1 y connectednetwork.



Figure 1 presents a simplified view of theorganization of the MERIT Computer Network.

The Michigan Interuni versi ty Committee onInformation Systems (MICIS) is made up ofrepresentatives from each of the threeparticipating universities. A few are computerexperts, but most are not. MICIS members in turnare responsible for selecting the Director of theMERIT Computer Network.

The MERIT Computer Network project director Isresponsible for the orderly execution of all of thetechnical and contractual responsibilities, broadlydivisible into three, staff supported functions!

1. Educational and promotional;

25

Joint Executive

Committee of

MERIT

1

I

I

1

Project

Director

Fiscal

Agent

Associate Directors

MSU U-M WSU NET Central

Senior Staff

NET Central

Figure 1. Simplified MERIT Computer Network Organization Chart

26


2. Research and technical development of thenetwork; and

3. Financial administration.

For the initial phase of network design andconstruction, MICIS chose a computer expert, Dr.Bertram Herzog, to lead the work. The director'soffice is located at the University of Michigan.

In the educational and promotional functions, thedirector is assisted by three associate directors,one located at each campus. The associatedirectors are chosen by the director fromnominations made by the universities. The specialresponsibility of each associate director is topromote and encourage the development of thenetwork within his own community of users. He hasesentially a dual role* to support the director andthe network so that the implementation of thenetwork at his particular site proceeds asefectively as possible, and to insure that his ownuniversity's interests are equitably served withrespect to the demands made on its resources by thenetwork. The associate directors have no directresponsibility for the technical details of theproject, but they are kept informed of relevantdevelopments and provide advice.

Four groups are concerned with the technicalresearch and development functions the NetworkCentral Staff, and the separate Network-MichiganState University, Network-University of Michigan,and Network-Wayne State University staffsindividually affiliated with their respective hostsites. Further, the Network Central Staff has twocomponents 2 the senior staff with technicalresponsibility for all of these groups, and theprogramming staff, charged with developing softwarefor the common part of the network. At the peak ofactivities, approximately the equivalent of 12full-time engineers and systems programmers wereinvolved in this facet of the project. This numberrepresents (equi valent ly ) two systems programmerslocated at each of the three campus computingcenters and six members of the Network CentralStaff located at MERIT headquarters.

Wayne State University acts as fiscal agent for the

27


network, fulfilling such functions as the receiptof funds from several sources, the distribution ofthese funds to the various MERIT groups andvendors, and the preparation of all contracturaland budgetary material.


1 . PI anni ng

Long range planning responsibility rests withMICIS, which continues to function as acommittee and currently meets on a bimonthlyschedule. Each university has appointed severalrepresentatives, officially four, to MICIS foran indefinite term. The representatives fromeach university consist of one high-rankingmember of each university's administration andof faculty members, usually including thecomputing center director and others Interestedin computer application areas.

In its efforts to develop a computer network,MICIS sought both state and federal support. Athree-member Joint Executive Committee wasestablished by the participating universities toadminister any funds provided by the State ofMichigan. Further, in the fall of 1966 theMichigan Educational Research and InformationTriad (MERIT) Inc. was created with Its threeman Board of Directors, and charged with theresponsibility to solicit and receive non-statefunds for the network. Wayne State Universitywas designated as the fiscal agent for all stateand non-state funds.

Once funding was assured, a network research anddevelopment project was established and workbegan in July 1969. Initially designated theTr i -Uni versi ty Computer Network, it later becameknown as the MERIT Computer Network Project.Thus the director of the MERIT Computer NetworkProject submits budgets for approval to theJoint Executive Committee and the Board ofDirectors of MERIT, Inc. (the same three peopleserve on both boards) whereupon the fiscal agentexecutes the appropriate contracturaloperat ions.

28


2. Instal 1 at ion

The special communications facilities for thenetwork were developed by the project staffunder the guidance of the Director. Thisdevelopment included communications computerhardware and common software. Software specificto a particular site, and all software residingin the host computers, is the primaryresponsibility of each host site's staff. Acommon design for network software was developedby the central project staff.

The Associate Director at each of theparticipating universities is responsible formatters relating to the installation of thenetwork facilities at his site.

3. Operations

Day-to-day management responsibility for thenetwork rests with the Director and his staff.

The MERIT staff is developing procedures toclosely monitor the performance of the network.Statistics gathered on message errors, trafficdistribution, and overall throughput willsignificantly help in adapting the originalnetwork design to actual usage patterns.

A. User Services

Responsibility for promoting the network withineach of the member universities and forproviding the required user services rest withthe associate directors.

The distribution of system documentationthroughout the user community is the jointresponsibility of MERIT and the individualuniversities. At the present time, MERITdisseminates information relevant to the designand operation of the communications subsystemand Its interfaces. Each university is requiredto maintain and distribute its local facility'sdocumentation and is responsible for issuingnotices reflecting any significant changes.

29



As explained below, the MERIT network was funded bythe Michigan State Legislature and the NationalScience Foundation. While this has not receivedmuch attention in the open literature, it is clearthat the network will need to be responsive to thewishes of the Legislature, Along these lines, Itis the intention of the network's developers toprovide the capability for network access by manyof the smaller colleges and universities in theState of Michigan without host computers.


A. Capitalization

The Merit Computer Network was funded, initiallyfor two years, by the Michigan State Legislature,the National Science Foundation and each of theparticipating universities. The Michigan StateLegislature provided in successive appropriationbills the total sum of $400,000, provided thatmatching support could be obtained from othersources. By the end of 1968, a proposal in theamount of $400,000 was submitted to andsubsequently funded by the National ScienceFoundation.

B. Accounting

The Director and the Joint Executive Committee ofMERIT have focused much attention on the accountingdifficulties encountered in even so controlled anetwork as this. The MERIT network follows thebasic policy of permitting each site to set pricesand charge for services individually. The problemis not in getting sites to offer resources forsale, but in convincing management at each of thesites to permit usage of resources at other sites.The problem Is that the possible "balance ofpayments" deficit (excess of outside use by localusers over inhouse use by outside users) cannot bepredicted in advance, and therefore is verydifficult to budget for, especially on a normalannual basis.

In MERIT'S case, relief was sought and obtainedfrom the universities' administrations to pledge,

30


from sources other than the computing budgets, anamount of monies to protect the potentiallyunbalanced budgets of the wholesaler (localuniversity computer centers) due to the presence ofthe network. By so doing, a deterrent toutilization was removed.

C. Regulation

The MICIS and the MERIT efforts have been formallyapproved by the Michigan State Legislature, theState Board of Education, and the governing boardsof each of the three universities.

Communications between sites in the MERIT ComputerNetwork currently utilize a pre-existing,inter-university, voice-grade telephone network.

D. Security

No special security procedures have been developedfor the MERIT network beyond the normal accesscontrol mechanisms for the individual host systems.However, all users of remote resources presentlymust be validated by their own system as well asthe remote system, since there is no access to thenetwork other than through a local host.

V. Conclusions

A. Summary of Problems

The developers of the MERIT network have had toface all the problems faced by developers of othernetworks of heterogeneous computers (e.g., ARPA)

.

The most publicized problems have been innon-technical areas. However, the development of aprocessor which offers a common interface to thenetwork from the specialized interface requirementsof each host machine was a non-trivial effort.Similarly, the development of the necessary controlsoftware for each of the host systems had to beundertaken. By and large the major technicaldifficulties have been overcome, although somepeculiarities remain to be ironed out. (Forexample, terminals behave according to the systemto which they are physically connected, not thesystem which they may be using, as In the ARPAnetwork).

31


Perhaps the most publicized non-technical issue inthe MERIT network has been the budgeting problemsgetting the computer centers of the threeuniversities to budget for possible net deficits innetwork usage (excess of work sent to other nodesover work taken in) for a period greater than asingle calendar year.

With this problem apparently solved (as explainedin IV-B), the main problem has been insufficientnetwork usage to justify its existence. It appearsthat the three participating universities each havesufficient computing resources to satisfy localneeds, so that there are few compelling reasons forextensive use of the network. Also, the doublecharges for use of remote resources (charges byboth the local and the remote system) tend todiscourage network usage. This problem could besolved by the expansion of the network to includeadditional user-only and direct terml nal -accesssites In the state, as the network developersenvisioned.

B. Lessons Learned

One important lesson to be learned from the MERITexperience is that organizational and managerialproblems can frequently dominate technical problemsin the development of a computer network, and caneven cast a shadow on technically sound networks.

In the technical area, MERIT has providedadditional evidence as to the soundness of using anintermediate communications computer to mediatebetween the standard operations of the network andthe peculiarities of each host.

Annexes

A. Bi bl iography

Aupperle, Eric M. "Merit network re-examined.", pp. 25-30.

Becher, William D. "The communications computerhardware of the MERIT computer network." , Vol. C0M-20, June 1972,pp. 516-526.

32


Cocanower , A. B. et . al. "The comrrunlcat I onscomputer operating system — the Initial design."MERIT Computer Network Manual 1070-TN-3, October1970 (PB 203 552).

Coleman, Susan (Ed.). "Users guide."(Preliminary). MERIT Computer Network publicationMCN-0473-UM-11, April 1973.

Eick, Harry, Seymour J. Wol fson and Karl L. Zinn."Facilities & resources available via the MERITcomputing centers." MERIT Computer Networkpublication MCN-0573-GE-U, May 1973.

Herzog, Bertram. "MERIT proposal summary." MERITComputer Network publication, February 1970.

Herzog, Bertram. "Computer networks." International Computina Symposium, 12-14 April1973, CINI Foundation, Venice, Italy, pp. 182-199.(MERIT Computer Network Publication MCN-0572-TP-8,May 1972.)

Herzog, Bertram. "Progress report for the periodJuly 1969-March 1971." MERIT Computer Networkpublication 0571-PR-4, May 1971.


CC - Communications Computer, the PDP-11 basedfront-end computer which provides a common networkInterface for host systems.

MERIT - Michigan Educational Research Triad

MICIS - Michigan Interuni versi ty Committee onInformation Systems

33

Triangle Universities Computation Center (TUCC)

I. Introduction

TUCC is an example of a successful cooperative venturein regional networking by independent and autonomousunl vers! ti es.



The Triangle Universities Computation Center is anot-for-profit corporation which is owned by thethree universities who cooperatively sponsored itsestablishment -- Duke University at Durham, NorthCarolina State University at Raleigh and theUniversity of North Carolina at Chapel Hill.


The network has three primary goals*

1. To provide each of the institutions withadequate computational facilities aseconomically as possible;

2. To minimize the number of systems programingpersonnel needed; and

3. To foster greater cooperation in the exchangeof systems, programs ano ideas among the threeuni versi t i es

.

Services include educational, research andadministrative computing services for the threemajor universities, about fifty smaller schools andseveral research laboratories.


TUCC is essentially a centralized, homogeneousnetwork comprising a central service node (IBM370/165), three primary job source nodes (IBM360/75, IBM 360/40, IBM 370/135), 23 secondary jobsource nodes (leased line Data 100s, UCC 1200s, IBM1130s, IBM 2780s, and leased and dial line IBM2770s) and about 125 tertiary job source nodes (64dial or leased lines for Teletype 33 ASRs, IBM1050s, UCC 1035s, etc.). Figures 1 and 2

34

DUKE/DURHAM

360/40

PRIMARY TERMINAL

50 Edtortranat IftstttatttiB-

UniversRies, CiHegis,

Community Colleges,

Technical Institutions,

& Secondary Schools

[Various Medium & Low

Speed Terminals)

UNC/CHAPEL HILL

360/75

PRIMARY TERMINAL

NCSU/RALEIGH

360/40

PRIMARY TERMINAL

Note: In addition to the primary terminal installation at DUKE,

UNC, and NCSU, each campus has an array of medium

and low-speed terminals directly connected to TUCC.

Figure 1. The TUCC Network

35

^

CO

r "o

P

01

zox3

z "^UJUJZ

uj oro-11-

53uj

Network Management Survey TUCC Network

Illustrate respectively the topology and geographyof the network.

Services to the TUCC user community include bothremote job entry (RJE) and Interactive processing.Thruput has grown from about 10,000 jobs per monthin 1967 to about 80,000 jobs per month in 1972.(This increase in thruput was accomplished byhardware upgrades during the period). At thepresent time about 8000 different individual usersare being served directly.


TUCC operates a 3-megabyte,telecommunications-oriented IBM 370/165 using0S/360-MVT/HASP and supporting a wide variety ofterminals (see Figure 3).

The three universities are connected to the TUCC bymeans of Telpak A (40.8 K baud) circuits whichconnect the universities' primary remote batchterminals to the central facility. In addition,over 50 educational institutions are linked to TUCCby a variety of medium and low speed lines whichcover the state and extend as far as ElizabethCity, Wilmington and Ashville.

All local node computers are of the samemanufacture as the central facility, and providelocal computing services and teleprocessingservices (from the central facility). Except inspecial circumstances, none of the local nodesprovide computing services for remote users at thistime, but the Hasp-to-Hasp software in use allowsfor such services to be provided if and whendesi red.

E. Future Plans

TUCC continues to serve its participatinginstitutions. Hardware and software adjustmentsare made as required to insure adequate capacity tomeet their processing demand. Immediate emphasisis on improving interactive services (primarily butnot exclusively via TS0) and services for largedata base processing including universityadministrative data processing.

37

to —— C3 "S

M-> LtJ °=:

=» 1

—

CO*x. •—o _j o

PO i-i 5Q_

C->

l_l_J -

i

^^




The network is characterized by an organizationwhich provides both for centralization of certainfunctions and the retention of freedom andauthority by the individual computing centers tooperate in the academic environment.

TUCC Organization

The TUCC Corporation is governed by a board ofdirectors whose nine members represent the threemajor universities. The three members from eachuniversity represent the administration,computer science instruction and computer users.The board members are appointed by the executiveofficers of each institution. The board meetsonce a month to act on matters of generalpolicy. Other attendees to board meetings arethe President and Director of TUCC, theAssociate Director and System Manager, theCampus Computation Center directors and theDirector of NCECS (see section C below). Mostquestions are decided by simple majority vote ofthe board, except that questions of "fundamentalimportance" are decided by each universitydelegation casting a single vote. Questions of"fundamental importance" include selection ofthe TUCC President, the annual budget and majorequipment decisions.

TUCC Staff

The central staff organization is shown infigure 4.

Systems Programming Section

This section is responsible for development,testing, integration and implementation ofall TUCC and manufacturers' system software.The section is headed by a systems managerwho is also the primary technical liaisonbetween the campus computation centers andthe manufacturers' field and systemsengineering organization. The systems

39

ASSOCIATE

DIRECTOR

PRESIDENT

& DIRECTOR

BUSINESS

MANAGER

SECRETARY

BOOKKEEPER

DATA BASE

PROCESSING &

OPERATING

SYSTEMS MGR.

SECRETARY

SYSTEMS

MANAGER

OPERATIONS

MANAGER

ASSISTANT

TO THE DIRECTOR

IPUBLICATIONS

ASSISTANT

INFORMATION

SERVICES EDITOR

SYSTEM

PROGRAMMERS

DEVELOPMENT

MANAGER

TELEPROCESSING

MANAGER

SHIFT

SUPERVISORS

TAPE

LIBRARIAN

KEYPUNCH ASST.

& CLERK-TYPIST

OPERATORS

Figure 4. TUCC Organization Chart

40


manager serves as an Associate Director ofTUCC.

Information Services Section

This group is responsible for the collectionand dissemination of documentation to users,campus center staff, directors and the Boardof Directors. Most documents are preparedthroughout the network organization,including the documentation services section.These documents are edited, approved andpublished by this section. The section isalso responsible for maintaining the programlibrary, for documentation standards, forpublic relations and visitor liaison, for aperiodic newsletter and for general interestbrochures.

Development Section

This group is concerned with generation ofnew versions of the operating system,maintenance of the manufacturer suppliedoperating system, designing and programmingof software interfaces between TUCC-writtenprograms and the operating system, andcreation of utility programs needed in theTUCC environment. The group is alsoresponsible for design, programming andinstallation of monitors, statisticsgathering programs for performanceevaluation, and for the evaluation of overallsystems performance.

Teleprocessing Section

The primary responsibility of this section ismaintenance of the teleprocessing software.It maintains current knowledge of allterminals and plans and acquires newcommunications hardware as required. Theteleprocessing manager also acts as aconsultant to the campus centers and toindividual users.

Operations Section

This section is concerned with the day-by-day

41


operations of the computer room. It alsomaintains the systems accounting records andprepares programs on systems usage,efficiency and turnaround statistics. Itprovides liaison with manufacturers' fieldengineers and with the campus computationcenter operations managers. It is alsoresponsible for security.


1 . PI anni ng

Once per month the campus computer centerdirectors meet with the TUCC director, systemsmanager, and assistant director, primarily todiscuss operational policies and procedures.Charging policies and changes to the billingalgorithm are among the topics determined at thedirectors' meeting.

The TUCC systems manager and the systemsprogrammers of the central staff and of theuniversities meet monthly to discuss plans fornew systems additions and/or modifications.

2. Instal 1 at i on

As described in 1 1 1. A above, all planning andcontrol over equipment installation at TUCC isperformed by the central management. Eachuniversity's computer center management performsthese functions for its own center.

3. Operations

Each university computer center is autonomousand is operated by its own staff. TUCC isoperated by a separate staff reporting to thecentral management.

4. User Services

Considerable attention has been devotedthroughout the TUCC organization to theprovision of adequate user services. Althoughthere is some information dissemination from thecentral organization in the form of regularnewsletters and an extensive documentation

42


system, the wholesaler-retailer organizationInsures that most user services are provided bythe local computer center. This facilitatesuser access to these services and insuresresponsiveness on the part of the providers.

In the case of the small institutions served byNCECS, these services are provided by "circuitriders'* who visit the schools on a regularbasis. Schools are visited as needed andgeographically close visits are usuallycoordinated. Consulting services are assistedby in-WATS telephone lines and personal contactsof remote users at workshops or through visitsto the central facility. Data lines can also beused for voice transmission when not otherwisein use. Some use is being made of inquiriesbeing sent by terminal communications to thecentral computer to be answered by the centralstaff.


The North Carolina Educational Computing Service(NCECS) was created as a state agency by the Boardof Higher Education in 1969. (NCECS is thesuccessor of the North Carolina ComputerOrientation Project (NCCOP) which began in 1966).Its mission is to provide educational services toinstitutions throughout the state. NCECS providestechnical support and consulting to small users.This includes computing services (terminals,communications and computer time) as well astechnical support (information services, technicalassistance to users, specialized software anddocumentation).

The director of NCECS represents his organizationat TUCC board meetings in a non-voting capacity,and also attends meetings of the computer centerdirectors. Close geographic co-location (indifferent wings of the same building) helpintercommunications between the two staffs,although the organizations are totally independent.

The NCECS staff includes both state supported andproject supported positions. Nine positions arestate supported* the director, his secretary,administrative assistant for curriculum

43


development, manager of user services, threecomputing consultants, an information servicesofficer and a business officer. Grant supportedpositions include curriculum development manager,programmer for curriculum development (half-time)and systems programmer (half-time).

The main function of the staff is the increase andimprovement of involvement of the participatingschools in computing. Two full time and onehalf-time "circuit riders" and a manager of userservices deal directly with the needs of theoutlying institutions, as explained above(III-B-4).


A. Capi tal i zation

Initial grants were received from NSF and from theNorth Carolina Board of Science and Technology, inwhose Research Triangle Park building TUCC waslocated. These funds, along with the payments forservices from the founding universities, served toestablish TUCC.

B. Accounting

The accounting system for TUCC is based on awholesaler-retailer concept. TUCC is a wholesalerof computing services, including machine cycles,operating system, programming languages andapplication programs, a documentation service, andmanagement. The TUCC wholesale servicespecifically does not include typical user services— debugging, contract programming, etc. — nordoes it include user level billing or curriculumdevelopment. Rather, these services are providedfor their constituents by the campus computationcenters and NCECS, which are retailers for the TUCCNetwork.

The wholesaler-retailer concept can also be seen inthe financial and service relationships. Every twoyears the founding universities negotiate with eachother and with TUCC to establish a minimumfinancial commitment from each to the net budgetedTUCC costs. Then, on an annual basis the foundinguniversities and TUCC negotiate to establish the

44


TUCC machine configuration, each university'scomputing resource share, and the cost to eachuniversity. This negotiation includes adoption ofan operating budget. Computing resource shares arestated as a percentage of the total resource eachday.

Each of the three universities and NCECS currentlypay 25% of the TUCC budgeted operating costs andare each entitled to equal amounts of service. Ascheduling algorithm with a "usage levelingcapability" allocates resources to the institutionwhich has used the least so far. Each institutionfunds its own computer facility and communicationslines. Each institution bills local users based onall of its costs, including payments made to TUCC,and on detailed usage statistics collected at thecentral computer.

The budget negotiation described above results inan effective wholesale rate to the threeuniversities which is a littie lower than thewholesale rate charged to NCECS. The justificationfor this procedure is the fact that the income fromthe universities is guaranteed while the incomefrom the NCECS is less certain. Both the computingcenters and the NCECS levy additional charges onthe local user to cover local computing centercosts and the costs of the additional NCECS centralstaff.

C. Regulation

No direct Federal or state regulations apply to theTUCC network. However, the state of North Carolinacan exert influence over the network through theUniversity of North Carolina.

Since the TUCC network does not extend outside thestate of North Carolina except for occasional dialservices, intrastate tariffs apply for mostcommunications facilities. Standard telephonecompany services are utilized for wideband andvoice grade circuits.

D. Security

In addition to those measures normally found in athird-generation operating system for the control

45


of access to files, TUCC has I mpl emented its owndata set and volume password systems as well aslocally developed encrypting programs and operatorprocedures.

V. Conclusions

A. Summary of Problems

Administrative Data Processing

TUCC has for some time been handling the fullrange of administrative data processing for twoNCECS universities and is beginning to do so forother NCECS schools. The primary reason thatthis application lags behind instructionalapplications In the NCECS schools is simply thatgrant support, which stimulated development ofthe instructional applications, has been absentfor administrative applications. However, thesuccess of the two pioneers has already begun tospread among the others.

With the three larger universities there Is agreater reluctance to shift their administrativedata processing to TUCC, although Duke hasalready accomplished this for their studentrecord processing. One problem which must beovercome to complete this evolution and allowthese universities to spend administrativecomputing funds on the more economic TUCCmachine is the administrators' (understandable)reluctance to give up a machine on which he canexercise direct priority pressure. The presentapproach to this problem is to extend theallocation and scheduling algorithm to guaranteea portion of the central machine to eachfounding university's administrative dataprocessing needs. Another problem Is thedevelopment of confidence in the availableback-up resources. It would probably requireadditional computing resources at TUCC if thisoption is elected by any of the universities.

Hardware Homogeneity

While not a real problem at present, it wouldappear that TUCC has locked itself into IBMcompatible systems. This has simplified the

46


development of the network by permittingcompatibility problems to be ignored, but it mayrestrict the alternatives for future growth.

B. Lessons Learned

1

.

User Servi ces

A very important lesson that was learned is thatpersonal communication must exist and be keptal i ve at al 1 1 evel s.

"It is amazing how misinformation can spreadif there does not exist a vigilant system forkeeping people informed... Experience hasshown us that if we relax ..., then littlethings that may go wrong may sometimes bemagnified completely out of proportion totheir importance and begin to become a sourceof irritation at some point in the system.The central facility must therefore have ahigh coefficient of sensitivity to the needsof all users. "(Davis, p. 4-1-2)

The earliest recognition of this fact was thehiring, at the time of the formation of TUCC, ofa Manager of Information Services at TUCC. Hisresponsibility is the documenting of alloperating systems, services and policies. Anelaborate system of memoranda series withdistributions to various relevant groups wasdeveloped. This lesson also explains the"circuit riders" who were employed by NCCOP (nowNCECS) to regularly assist its client colleges.

2. Wholesaler-Retailer Concept

TUCC's implementation of the wholesaler-retailerconcept (as explained In section IV-above) wasdesigned as a mechanism for the administrativeprotection of the interests of the threefounding universities and the NCECS schools.

Because each of the universities and NCECS isguaranteed a minimum percentage of utilizationof the central machine (in effect, a virtualmachine for each), they have the assurance thatthey can take care of their users' needs aswould be the case with totally independent

47


facilities. The scheduling algorithm alsoallows each to define and administer quiteflexible, independent priority schemes.

Since the local centers and NCECS are theretailers of all computer services, whetherproduced locally or purchased on a wholesalebasic from TUCC, they are not in competitionwith TUCC. Users are also able to turn to localpersonnel for all required services, and receivea si ngl e bi 1 1 .

There are several structural devices which serveto protect the interests of both the wholesalerand the retailers. At the policy making levelthis protection Is afforded by the Board ofDirectors, which is appointed by the Chancellorsof the three founding universities. Typicallyeach university allocates its representatives toinclude its business interests, its computerscience instructional interests, and its othercomputer user interests. The UniversityComputer Center Directors sit with the boardwhether or not they are members, as do theDirector of NCECS and the President of TUCC. Anexample of the policy level function of theBoard is their determination, based on TUCCmanagement recommendations, of computing servicerates for NCECS and other TUCC users. (Williams,1972)

At the operational level there are two importantgroups, both normally meeting each month. TheCampus Computation Center Directors* meetingincludes the indicated people plus the Directorof NCECS and the President, the Systems Manager,and the Assistant to the Director of TUCC. TheSystem Programmers* meeting includesrepresentatives of the three universities, NCECSand TUCC. In addition, each of the universitieshas the usual campus computing committees.

3. Neutrality of Site

The neutral location of the central computer isfelt to be "one of the chief factorscontributing to the political stability ofTUCC." (Brooks, et. al.) An earlierunsuccessful experience with a computer jointly

48


owned by NCSU and UNC-CH, but located at ChapelHill, had shown that "the psychological andpolitical consequences of location could not betele-processed away." It is recognized that aneutral location requires extra cost, but thisis felt to be "an i ndespensi ble expense."

Annexes

A. Bi bl iography

Brooks, Frederick. P., Jr., Ferrell, James K. andThomas M. Gal lie. "Organizational, financial andpolitical aspects of a three-university computingcenter." Proceedings IFIP Congress 1968>, pp.923-927.

Davis, M. S. "Economics -- point of view ofdesigner and operator." (Austin, Texas, April 20-22, 1970), University ofAustin and The Mitre Corporation, pp. 4.1.1-4.1.7.

Freeman, David N. and Robert R. Pearson."Efficiency versus responsiveness in amultiple-services computer facility." , 1968, pp. 25-34.

Freeman, David N. and Joe R. Ragland. "Theresponse-efficiency tradeoff in amultiple-university system." , Vol.16, No. 3, (March 1970), pp. 112-113.

Grobstein, David L. and Ronand P. Uhlig. "Awholesale retail concept for computer networkmanagement." 1972 FJCC, pp. 889-898.

International Business Machines, Inc. "In NorthCarolina, computing power for education.", 9 : 2 (Summer 1973), pp. 2-6.

Parker, L. T., Gal lie, T. M. , Brooks, F. P. and J.K. Ferrell. "Introducing computing to smallercolleges and universities — a progress report.", 12:6 (June 1969), pp. 319-323.

Williams, Lei and H. "A functioning computernetwork for higher education in North Carolina."1972 FJCC, pp. 899-904.

49



NCCOP - North Carolina Computer Orientation Project

NCECS - North Carolina Educational ComputingServi ce

TUCC - Triangle Universities Computation Center

50

Oregon State Regional Network

I. Introduction

The Oregon State Regional Network is representative ofthe large number of centralized "star" networksserving a regional clientele. It was organized toprovide educational computing services to a number ofgeographically dispersed institutions, and is notconcerned with research into the technology ofnetworking,



The network was sponsored by Oregon StateUniversity with support from the National ScienceFoundati on.


The regional network was established in connectionwith a two-year project with National ScienceFoundation support to develop and appraiseinstructional uses of computational facilitiesprovided through computer terminals on-line to acentral computer facility. The objectives of theproject were*

1. To provide faculty and students at theparticipating colleges with computing resourcesavailable through terminals which have directaccess to a central computing facility.

2. To develop special instructional programs andmaterials suitable for regular as well as"short" courses in the use of computers in aneducational environment.

3. To appraise the usefulness in instructionalprograms of the facilities offered withreference to the instructional needs of eachinstitution.


The system now supports a network of more than 200remote terminals with approximately 75 terminalsactive concurrently. Approximately 35 terminals

51

Network Management Survey Oregon State Regional Network

are located at other universities and colleges ofthe network. The network serves instructional,administrative and research aplications, and may bereaching saturation. Acquisition of new computinghardware ?s being contemplated at the present time.


The Oregon State University Regional ComputerNetwork consists of a central computing facility atthe Oregon State University campus in Corwallls,Oregon, which serves some sixteen institutions ofhigher learning in Oregon. The central facilityruns in a time-sharing mode under a specialoperating system developed at Oregon StateUniversity. Remote users are connected to thecentral facility by low speed teletypewriter lines.

The principal resource at the computation center isa CDC 3300 with a memory capacity of 98K 24-bitwords. Peripheral devices include a card readerand punch, line printer, four magnetic tape units,five magnetic disc units, a 200 million charactermass storage disc unit, 16 CRTs, a plotter and aPDP-8 minicomputer which serves as a communicationsfront end. The PDP-8 also serves as a telephoneline interface for 180 remote terminals. During1972 remote batch capability was added using cardreaders and line printers.

The operating system, OS-3 (Oregon State Open ShopOperating Syste^ was designed and implemented bythe Oregon Stat

Network Management Survey Oregon State Regional Network

network, with a central facility and staff at OSUand remote stations at the other participatingsi tes.


A, Structure and Extent

The organization of the network is embedded in theorganization of the regional computer center.Special organizational elements, exclusivelyconcerned with network operations are a regionalcoordinator who reports to the regional computingcenter director

Network management survey - NIST...6."NetworkManagementSurvey" I.W.Cotton February,1974...

Documents

Transcript of Network management survey - NIST...6."NetworkManagementSurvey" I.W.Cotton February,1974...