Internetworking: an analysis and proposal

Theses Thesis/Dissertation Collections

1990

Internetworking: an analysis and proposal

Gary M. Diana

Follow this and additional works at:http://scholarworks.rit.edu/theses

This Thesis is brought to you for free and open access by the Thesis/Dissertation Collections at RIT Scholar Works. It has been accepted for inclusion in Theses by an authorized administrator of RIT Scholar Works. For more information, please [email protected].

Recommended Citation

Gary M. Diana April 26, 1990

Rochester Institute of Technology Department of Computer Science A Thesis, submitted to the Faculty of the

School of Computer Science and Information Technology, in partial fulfillment for the Degree of

Master of Science in Computer Science

Approved by:

~/(

Ie ()

Associate Professor Dr. AndrewT. Kitchen Chairman, Thesis Committee

Associate Professor Dr. James E. Heliotis Reading Member, Thesis Committee

I Gary M. Diana hereby grant permission to the Wallace

Memorial Library, of RIT, to reproduce my thesis in whole or in part. Any reproduction will not be used for commercial use or profit.

This thesis represents the final efforts on _{my way} to _obtaining the degree ofMaster of Science in Computer Science. There are _many people who supported me over the six yearsit_{has taken,} andIwouldliketo take this_opportunitytorecognizethem.

I wouldlike to first thankDr. Ronald Sarnerofthe SUNY Institute of_Technology and

Dr.Leonard SmithoftheSUNY CollegeofEnvironmental Scienceand_Forestryfor

then-letters of recommendation into the Master's program. _They were instrumental to _my

admission. Iwould alsoliketo thankEastman Kodak_{Company, specifically}theSystems

Solution_Group within Business _Imaging Systems. _Theynot_onlyprovidedrelease time

and tuition_support, butalso a resource-rich environmentin whichto_carry out_my work.

Rich _Bartl, Dan _Himes, and _Mary Alice _Flagler, of the Eastman Kodak Technical Library, assisted_{greatly in} theprocurement of research materials. _They saved me _many hourstime.

I would like to thank _my thesis committee fortheir assistance. Dr. Andrew T.Kitchen,

Dr. James E. _Heliotis, and_HenryA. Etlingerprovidedtechnical,professional, andmoral

support. _They made what could have been a _long drawn-out process an enjoyable experience. _Theywere_especiallypatientwiththe_earlyrevisions of_mywork.

Iwouldlike to thank_myfamily andfriendsfor theirsupport. _Theyall_{listened patiently}

to complaints oftoo much work to _do, programs thatdidn't work, and due dates. _My parents,LouisandMarilyn _Diana, made numeroustrips to Rochesterto _babysit, whileI didschoolworkand_mywifedid her nursingwork. I simplywouldnothave beenableto

complete the coursework withouttheir help. I would also liketo thank Davidand _Jim,

fortheirinterestand supportinthisendeavor.

Finally, Iwould like to thank _mywife Sheryl and our son _Gary Jr. While I was _taking classes and_{during mylong} hours in the grad_lab, she managed a_career, a _house, and a

typeAtoddler. There isno oneelse whowouldhave donethisforme. _GaryJr. was_very patient with aDadwhodidn't alwayslet him playwithhiscomputer, orhave the time to

playallthegamesthreeyearoldslike toplay. Ican nowlookforwardto tee_{ball, fishing,}

As the number of computer networks has grown, so has the desire for users on these

networks to communicate with each _other, thus the need for internetworking.

Unfortunately, many of these networks were not designed with _{internetworking} capabilities in mind. The _{internetworking} facilities offered _by a typical networkrange

fromnon-existentto state oftheart. Twomajor efforts towards _{internetworking} arethe

DARPA Internet protocols and the OSI _{Internetworking}protocols. The goals ofthis thesisaretoacquaintthereaderwiththequalitieswhicharedesired in an_{internetworking} scheme, to describe how _{internetworking} is accomplished currently, and how these

protocols mightbemodifiedtobettersuittheneedsoftheinternetworkuser. Tothisend, this thesiswill _develop thefunctionalrequirementsforan

"ideal"

internetwork, describe

two current methods for _{internetworking,} and analyze these methods against the ideal internetwork. Theadvantages anddisadvantagesof each_{internetworking}methodwillbe

discussed. Afterthisanalysis, suggestionswillbemadeas tohowthese_{internetworking}

Chapter 2 Problem Descriptioa 4

Chapter 3

-ThesisLimitations 12

Chapter4

-The OSI Approachto_{Internetworking} 13

Sect. 4.1 TheOSIStandardsfor_{Internetworking} 17

Sect. 4.2- TheNetwork

LayerArchitecture 18

Sect.4.3

-General Network LayerServices 20

Sect.4.4- OSI

InternetworkingStrategies 22

Sect. 4.5- Connection-oriented

Network Service_(CONS) 24

Sect.4.6

-Connectionless Mode Network Service_(CLNS) 41

Sect.4.7

-Protocolsfor_ProvidingtheOSI Network Service 44 Sect. 4.8 AConnection-modeNetwork Protocol_(X.25) 45

Sect. 4.9- AConnectionlessNetwork Protocol (OSI

IP) 58

Sect. 4.10-TransportLayerServices 71

Sect. 4.11 TransportLayer Protocols 78

Sect. 4.12- Analysis

ofOSI_{Internetworking}Standards 80

Chapter 5 - The DARPA Internet

85

Sect. 5.1 The InternetServicesandProtocols. 86 Sect.5.2

-The Internet Protocol_(IP) 87

Sect. 5.3 - The Internet Control Message Protocol

(ICMP) 92

Sect.5.4- The Transmission Control Protocol

(TCP) 96

Sect. 5.5- The

Gateway-to-GatewayProtocol_(GGP) 102

Sect. 5.6- The Exterior

GatewayProtocol_(EGP) 104

Sect. 5.7 The Interior_GatewayProtocol_(IGP) 107

Sect. 5.8-AnalysisoftheDARPA Internet 109

Chapter6- An

InternetworkingProblem andProposal 114

Chapter 7 Epilogue 123

Chapter8 Literature Search 128

Chapter9

-Glossary 129

Chapter 10

-Bibliography 133

Currendytheworld isfull of computer_{networks, mainly workingindependently} ofeach

other. This is due to several factors. Often thereis no company-wide _{strategy in} place

for the purchase of similar computer systems which can be networked together. This

results in "islands" ofinformation withinacompany, with little meansfortheexchange

ofinformation. Some computer systems run _proprietary networksoftware andthis can

present _connectivity and _{interoperability} problems. And _lastly, physical location of a

given computer network can lead to isolation ofit from other computer networks and

systems [12,20]. Asowners of systems become more and moreinterested in sharing the

resources of other _networks, the need for interfaces between networks becomes more

urgent. _{Internetworking}was often an _{afterthought,}if itwasthoughtofat all. As aresult, because of their inherent _{differences, many} networks do not _{easily interface} without

additional functionality. This _{functionality} can be present in the form of a specialized

gateway computer. The gateway would be responsible for _interfacing to _potentially different physical media as as well as _converting the protocols used _by the _differing

systems.

Examples ofthevariousdifferences betweennetworksinclude:

*

thephysical mediumusedto_carrytheinformation.

theformatofthedatacarried onthephysicalconnection.

themethod_(protocol) for exchangingthedata.

the_namingconventions usedto_identifyothercomputers and networks.

the_qualityof servicesoffered_bythenetwork.

routingmethods

Tocope withthesedifferences, various schemeshave been devised.

"Bridges"

have been

designed to span networks which have differing physicallink layers.

been designed to span networks which have _differing data link layers. "Protocol

converters"

havedesignedtospan networkswhichhave_differingprotocolsin theupper4

layers oftheOSIreference model. None ofthese solutions are perfect. Their tradeoffs,

advantages, and_{disadvantages}willbe discussedin thecourseofthis thesis.

There aretwomajor groups_addressingtheissueofinternetworking. Oneofthesegroups

is theDefense Advanced Research Project_Agency (DARPA). It hasdeveloped whatis

currently known astheDARPAInternet. Thisinternet hasbeen inexistenceforoverten years. It has changed and evolved over the years, as a result of actual use and

experiences. The other _group involved in _{internetworking} efforts is the International Standardization Organization (ISO). _{This is only} part of their _{responsibility,} which

includesthe specificationofarchitecture and protocolswhich span all sevenlayers ofthe

OSInetwork model.

The DARPA Internet has made a large impact in the area of internetworking. The Internetconsists of_{approximately 2400hosts,}400networks, and_manytimes thatnumber

of users [47,58]. "Requirements for Internet _Gateways", RFC 985 _[58], estimates the

Internetgrowth rate at 10% permonth. Theprotocols used_bytheInternetare_commonly

knownasTCP/IP. The IP (InternetProtocol) isthenetworklevelprotocol responsiblefor

interfacing to the datalink_layer, and_controlling the_routing ofdatapackets. The TCP (Transmission Control _Protocol) is responsible for _interfacing to the IP and offers the

services of reliable data transfer andflowcontrol. TheInternet protocols are attractive

for manyreasons. One is that _they have demonstrated their _ability to perform_reliably

andefficiently, andareconsideredtobematureand stable [61].

TheOSIprotocols are notwithout merit. Theymay hold thegreatest eventual potential

for_linkingvarious networksinto the world'slargest internet. However, theOSI design

Theproposedthesiswillhavethe_followinggoals:

To define an "ideal" _{internetwork, in} order toanalyze _{internetworking} protocols

andservices.

To describe how _{internetworking} is handled _by the DARPA Internet and OSI

protocols.

TocomparetheDARPAandOSI_{internetworking}to the "ideal"internet.

Topropose a new set of_{internetworking protocols,}orenhance_existingprotocols

and services.

Throughout this _thesis, the Open Systems Interconnect _(OSI) Reference Model will be

referenced. Thisisthewell-known sevenlayermodel,which specifies adifferent levelof functionality for each layer of the model. This model can be applied to computer

systems,anditslowerthreelayers are concernedwithvariousfunctions whichcontribute

to internetworking. Where appropriate, a given protocol analysis _{may be}mappedonto

the sevenlayermodel. Thisis done inorderthat thereader_maygain an_{understanding}of

howoneprotocolcompareswithanother.

In order to fulfill the first goal of this thesis, an internetwork will be defined. An

internetwork, or_internet, can bedescribedas a network of networks. Itis a result ofthe

interconnection of two or more networks, often for the purpose of greater resource

sharing,communications,and/ortoprovidea singleadministrationpoint. To separatethe

notion of a single networkfromthatofan_{internetwork,}singlenetworks willbe referred

to as subnetworks. It is therefore the case that internetworks are the result of interconnectingone or moresubnetworks.

An "ideal" internetwillnowbedefined. This definitionwillbeafunctionalspecification

of whatauserwould_ideallywant aninternetwork tooffer. Itshouldberealizedthatsuch

represent astandard_bywhich otherinternetworkscan be analyzed. Since differentusers desire different_{services,fitness} as a"good"internetwill bedependent to some extent on

theusage.

Thequalities of an"ideal"_{internetworkingdesignincludethe}

abilityto: be_easilyextended.

interfacetoas_manydifferent subnetworks as possible. function inthefaceofnode/subnetwork/internetworkfailure.

maintain alevelof performance acceptabletosubnetworkusers. *

provide variouslevelsofdatasecurity.

maintaindata integrity.

hidethe specificsofthesubnetworkinterface fromtheuser.

allowfordecentralizedadministration. handlethe_accountingoftraffic (data).

The qualities of an ideal internet were derived from several sources [5,9,13,41,43,65,68,70,72]. After the ideal internetwork is defined, two well-known

internetworking solutionswillbe examined. The_{internetworking}schemes examinedwill be evaluated on the basis of the above criteria. From the _analysis, weaknesses and

strengths will be identified for each internetwork. Conclusions and recommendations

willbemade onthebasisoftheseanalyses.

The DARPA andOSI_{internetworking}protocolswillbeexamined as a second goal ofthis

thesis. _{Internetworking} has been extensively researched in the past 20 years. The implementation ofARPANET inthe late sixtiesis a good example ofthis. This isareal

life, working example of an _{internetworking} solution. Over the last ten _years, the

DARPAInternethasgrown_{extremely large. It}consists ofhundredsof_machines, groups

ofwhich _belong to various kinds ofnetworks. For these reasons and because of the

according tothe idealinternetwork. The Internetprotocolshave evolved overthe years, andthe_historyofthemandtheir_{design philosophy}willbe discussed.

The other_{internetworking} scheme to be examined isthat offered _bythe OSIprotocols.

The OSI architecture, services, and protocols have specifiedforeach layerofthe seven layerOSI Reference model. Although some ofthese protocols are still _beingdiscussed

bycommittees, thereis enough informationavailableforan analysis. Theprotocols and

architecture specifiedforthe transport and networklayersare thosewhich have themost

impacton_{internetworking}design. Thesolution offered_by theOSIprotocolshasperhaps

thegreatestpotentialfora _followinglargerthan_anyother_{internetworking}design. Major

computer_{vendors, third-party} software _houses, theU.S. _government, and others have

already beguntoline_{up in favor}ofit. The U.S. governmenthas statedthat all software purchased in the near future must conform to the OSI protocols [69]. However, not

everyone is a proponent ofthe OSI protocols. There are critics that claim the OSI

committees did not give enough weight to the experiences learned in the DARPA

community. Due to the _{potentially large impact}ofthe OSI _{internetworking}protocols,

andbecause theeffort is notwithout critics,it toowill be analyzed and comparedto the

ideal internetwork.

Finally, this thesiswillproposeitsown_{internetworking}design. Itwillbethegoal ofthis

design to fit the specifications ofthe ideal internetwork in a _way that is superior than

either DARPA or OSI solutions. It is not expected that this design will be _radically different from either the DARPA or OSIprotocols, since one ofthe goals ofthe ideal

internetis the_abilitytointerface to_existing subnetworksandinternets. In_fact,it_{may be}

that the proposed internet design is an enhanced version of one ofthe aforementioned

TheIdealInternet

This section will discuss the qualities and characteristics of an "ideal" internet. Each

qualityoftheideal internetwillbeoudinedfromauserand/orinternetworkadministrator

perspective. The_following are qualitiesthatanidealinternetwouldpossess:

Extensibility

Flexibility

Interoperability

Robustness

HighPerformance

Data_Integrity

Transparency

EaseofAabrrinistration

Traffic_{Accountability}

Extensibility

Extensibilitycanbedescribedasthe _abilityto_easilyextendtheinternet. The additionof

new_{subnetworks, gateways,}and nodeswillbe hampered iftheinternetrequires elaborate

steps to be followed. It is expected that _any new subnetwork added on will have to

register with one or more central controlpoints. This is at the _{very least necessary for}

reasons of_routing and addressing. Given human nature, if it is difficult to add new members to the internet, few additions will be made, and this will limit the ultimate growth and_usabilityofit.

Flexibility

This quality of an internet can be described as the _ability of the internet to change as

technology changes. For example, if a new physical medium _technology becomes

available and it isdesirable from auser'sperspective, then theinternet shouldbeableto

configuretheir subnetwork. The users candecidewhere _any given _{functionality} should

reside, based on cost and performance. _Following the OSI reference model is one methodfor_{partitioningfunctionality, regardless ofthevehicle usedfor executing it. To} do less than this would almost ensure_{obsolescence,} or atthe_{very least limit}the growth

potential of the internet. It is more important to _{design for easy} modification ofthe internetwork, than to put together a single optimized network design [9], It is also

important that the internet be able to_employ new protocols as _they become available.

This_{is especially necessary if}thebandwidthofthephysical mediumis_increased,beyond

the limits of the higherlevel protocols. Another aspect of _flexibility is in the type of

services offered _by the internet. For_example, auser _maywantto emphasize thequick delivery ofdataversus the _accuracyofthe data. Such isthecase with digitized speech.

Inother casestheuser_may wishto emphasizethe _accuracyofthedata,as opposedto the

speed of_{delivery. This is usually}thecasewith file transfers. Thepoint here isthatitis desirable toallowtheuserthe_abilitytochoosetheservice_required,nottomandateit.

Interoperability

This quality of an internet can be described as the _ability to interface to a _variety of

heterogeneous hosts and subnetworks. This quality becomes _increasingly important as

theneeds ofthe userrequirethe _sharingofexpensive_resources,which resideonremote

hosts and subnetworks. The _difficulty with _implementingthis feature of an internet is

thatoften asubnetwork vendormakesno attemptsintheinitial designof a subnetworkto

include the _ability to interface withforeign subnetworks. Currently, _many vendors are

moving toward the support ofde facto standards such as TCP/IP. It canbe envisioned

thatifallvendors_{already had}compatible protocolsinplace,there wouldbe littleneedto discuss thisissue.

Robustness

gateway,subnetwork, or_{internetworkfailure. Nodefailure is inevitable,} and atbestcan

only be minimized. Even ifa given piece of hardware is_{only inactive for}preventative

maintenance, thesegapsin_availabilityneedtobeaddressed. Fromtheuser'sperspective, it should_{be unnecessary} todetectand re-establish communicationwith a remote nodeif

an _{intermediate gateway} or subnetwork fails or otherwise becomes unavailable. The

internetworkingscheme shouldhandlethis. There are various methodsfor providingthis

service, and the implementation can affect _performance, and has a direct effect on the amountofstateinformation thatneedstobe keptattheintermediatenodes.

HighPerformance

The performance of an ideal internet will not _drop below the level users on the

subnetwork expect. The actual performance realizedwill be dependent on theinternet hardware and _protocols, as well as the performance level of _{any intermediate nodes,}

gateways, and subnetworks. The ideal internet will allow _delivery of data at a rate sufficient for even the most _data-heavy applications. An example of a _data-heavy

application wouldbeone_involving thereal-time_delivery ofdigitized speech. Real-time digitized speech is _interesting since late data is worse than incorrect data. These

requirements for data _delivery rate aremuch more stringentthan those for file transfer. File transfernecessitates an accurate_deliverysystem, as opposedtoa quick one.

Data_Integrity

To someinternetwork serviceusers, dataintegrity isanimportantquality. Forthose that

advertise data _integrity as a _{feature, they} are obligated to insure the _integrity of transmitteddata. Itshouldbenotedthatsome_{subnetworking}protocolsdonotinsure data

integrity (ordelivery), andleave dataintegrity _up to the higher layers (i.e. the transport layer). Data _integrity can be accomplished on a _{hop by hop} basis, or on an end toend

Transparency

Transparencyisthe_abilityoftheinternettobeextendedin a_way thatisuneventful to the

rest oftheinternet. _{For example, it}would notbeappropriatefortheentireinternettoshut downfortheaddition of a new _node, orin ordertoinstall new software on a node. Part

of_transparency is related to the robustness of the _internet, i.e. how does the internet

handlea situation when a nodeis_comingup,_goingdown,or_simplynot_handlinginternet

traffic at a given momentin time. In _summary, the user on the internet should not be disturbed, or have his activities affected _{simply due} to some modifications _going on

elsewhereintheinternet. _{Another way}to lookat_transparencyis fromtheperspective of the actual data as it travels through the internet. It is quite possible that a given data

packet will be _fragmented, or _{broken up} into smaller data packets as it travels from subnetworktosubnetwork. This is dueto thefactthatall networksdonothavethesame

maximum size for a data packet. This necessitates that the packet be at some point brokenup, thenreassembled priorto_deliveryto the destination. Theuser atthe source

(or _destination) node should not be concerned with this fragmentation or _reassembly

process. The idealinternetwillexhibit ahighdegreeoftransparency.

Administrability

This qualityof aninternetaddressesthe_abilityofinternettobe administered. Aninternet

canbe administeredintwo ways. Onemethod of administrationis foronecentral_group

(agency, university, research _lab, company, or_country) to administertheentire internet.

This is the centralized approach. This has the advantage of a single focal point

controlling the administrative functions. There would be no arguments or agreements

necessary; whatever the central administration did, would go. This has the distinct

disadvantage that there are_probably no groups that wouldcare to have control of

then-subnetwork relinquished to an external group, especially one in another country. The

other method ofinternet administration is a decentralized one. This method involves

can administeritself_{any way} it wants to. Ofcourse, there are certain functions which willrequirecoordinationbetween sitesifa given subnetworkisto_belongtotheinternet.

There are several facilities which anideal internetadministration package would allow

for. _First,itwould allowfor_debuggingoffaultswhicharedetected. Thiswouldinclude

functionality toreport on faults as _{they occur,} a methodfor _loggingthese_reports, and a

method forthe execution ofdiagnostic tests [39]. Another part ofadministration of an internet is the _ability toconfigure a given subnet, or domain. This _{includes naming} of

nodesin thelocal subnet,reconfiguring the_topology, and_{communicating}thesechanges backto theinternet.

Traffic_{Accountability}

This quality of an internet is the _ability to charge a given user for the passage ofdata

through a subnet and/or for the useofservices. This information wouldhave sufficient

granularity so that itcouldbe determined how muchtobill agiven user. This wouldbe

Chapter 3 Thesis Limitations

This thesiswill not attemptto prove_any oftheprotocols it suggests. That _is, there will be no_{implementation} to_{test, in} actual _use, the enhancements suggested. _However, for

any enhancements, there will be a desciption ofhow it would operate, along with _any advantages and disadvantages. An implementation of a suggested protocol could be

accomplished as a separatethesiseffort.

There will be no attempt made to examine all the schemes which _currently existto do

internetworking, although it is recognized that others do exist. The DARPA and OSI

protocolshave been chosenbecause oftheircurrent widespread use and potential future

use, respectively. Examples of otherinternetworks include JANET, ROSE, the Xerox

Internet, theAT&T_Internet, DECEasynet, CAMPUSNET _[11],NRI (National Research

Internet), andBITNET.

The NRI is a collectionofFederal agency-owned subnetworks. The AT&T Internetis

administered and used_{internally byAT&T;} it isTCP/IPbased. DEC Easynet is basedon

DECnet, and is used _{internally by} Digital Equipment Corporation. JANET isthe Joint

Academic NETwork, based in the United Kingdom. ROSE is the Research Open

Systems forEurope; it is_beingusedas a_provinggroundforthe ISOprotocols. BITNET

is the Because It's Time NETwork; it consists _mainlyof_University members. A brief

Chapter4 The OSIApproach to_{Internetworking}

Before_attemptingto describethemethodsfor_{internetworking} developed _byISO, abrief

history leading to their evolution will be given. The _{ISO, seeing} a needfor a unified

approach for computer _{communications, developed} what is _{commonly known} as the

seven layer OSI reference model. This model partitions the task of _{interconnecting} computer systems into seven distinct and well-defined layers. The ISO standards

documents specifythismodel_fully [25]. From highesttolowest_layer,thelayers defined

bythereference model are:

Application

Presentation

Session

Transport

Network

Data Link

Physical

Application Layer

This layer supplies those services _commonly needed _by the user applications layer

services. This _typically can include such things as virtual _{terminal, file transfer,} and electronicmailservices.

Presentation Layer

This layer supplies services which are required _{by sufficiently many applications,} that a

general solution is indicated. Examples of this include data representation and

compression. Takeforinstance,theproblemof_{exchanging integer}dataon end machines

Session Layer

This layer allows end systems to establish sessions with each other. This is usefulfor

synchronizing a _dialogue between the two systems. For example, where there is a

half-duplex connectionbetween end_systems, thesessionlayerservices could_keeptrack

of whose turn it is to transfer data on link. Synchronization services _may also include

remembering where atransaction left offwhen a crash occurred. This is useful when

doing a_lengthy filetransfer. This_{feature, commonly known}as _{"checkpointing",}would

allow the transfer toresume at_{approximately}thepoint ofthelastcrash.

TransportLayer

The services offered_by this layer includethetransportation ofdata from end system to

end system. Ifneed _be,thedatacanbe segmented_bythis layer andlaterreassembled at

the destination end system. The transport layer can also do multiplexing, in order to

speed_{up delivery} and increase throughput. This all occurs _{transparently} to the session

layer. Thisisalso thelowest layer in themodelthat isconsideredtobe involved in

end-to-end communications with its peer entity. That _is, the _{layers beneath it only}

communicate with the next intermediate system, which _may not _{necessarily be} the

destination end system.

Network Layer

This layer provides high level access and control of the subnetwork upon which it

operates. This is the layer responsible for the _routingofdata packets. _{This layer may}

also provide congestion _control, accounting, and communication facilities over

homo/heterogeneous subnetworks. As willbe shown _later, this layer has its own

mini-architecture, consisting of three sublayers. This is due in part to the _diversity and

complexity of the various subnetworks and the methods which can be used to

DataLinkLayer

This layerprovides an errorfreecommunication service across a given subnetwork. To

accomplish _{this, it} takes a primitive physical transmission _facility (offered _by the

physical _layer) and transforms it into a reliable transfer medium. This occurs

transparendytothenetworklayer.

Physical Layer

The physicallayerprovides the servicefor sending rawbits over thephysical medium.

This layerdealswiththeissuesof_{procedural, electrical,} andmechanicalinterfacesto the

physicaltransmissionmedium[68].

The OSI reference _model, like all ISO _standards, was discussed and agreed upon _by

international committee. The seven _{layer partitioning} came about _{by following} these

generaldesignguidelines[40]:

theoverallarchitecture shouldbesimple keepinterfaces between layerssimple

functionswhich are_{very differentbelong}in separatelayers

combineintothesamelayerthosefunctionswhichare similarinnature

usepast experiencetoselect_boundarypointsbetween layers

maximize _{ability for}change _bycreating layers whereinternal mechanization can bemade_{independently}fromthe_{functionality}

createlayer boundarieswheretherequirement_mayexistin thefuture createlayerswheredifferent levelsofdataabstractionare required

createlayer boundarieswith_onlythelayerabove andthelayer below

allow_sublayering tooccurwherecommunicationservices requireit

The_{followingdiagram illustrates}theposition of each layerwithrespectsto theothers:

APPLICATION APPLICATION

PRESENTATION PRESENTATION

SESSION SESSION

TRANSPORT TRANSPORT

NETWORK NETWORK

DATA LINK """ms DATALINK

PHYSICAL PHYSICAL

The dotted lines indicate a logical connection between the _layers, while a solid line

indicatesaphysical connection betweenthelayers. Eachsevenlayer

"stack"

ismeantto

Sect.4.1 The OSI Standardsfor_{Internetworking}

This section will summarize theOSIstandards specified _byinternational committeethus

far that _{involve internetworking. The section following this will detail the} services

offered_bythenetwork andtransportlayers. Exampleswillbeprovided ofhowa service is derived from an actual protocol. The options open to an OSI implementor are more

diverse thanthoseoffered_bytheDARPAprotocols. Themaindifference isthattheOSI

protocols allow both connection-mode and connectionless-mode networklevel services. With TCP/IP _{internetworking,} the network service _{is strictly} connectionless. Thus the

descriptionoftheTCP/IPnetwork servicesis a more straightforward one.

In order todescribe the services and provisionsforthese services atthe network _layer,

severalOSIstandardshave been issued:

ISO 8648 describes theinternalorganizationofthenetworklayer. This document

detailsthe threelayermini-architecturethatexiststhere.

ISO8348 describes theconnection-mode network servicedefinition.

ISO 8348/AD1 (Addendum ₁₎describestheconnectionless-mode network service

definition.

ISO 8348/AD2 describesthenetwork_{layer addressing}details.

ISO 8880 Part 1 describes the generalprinciples and conformance _{for providing}

thenetworklayerservice.

ISO8880Part 2 describestheprovisionsfor supportingconnection-mode network

service.

ISO 8880 Part3 describes theprovisionsfor supporting theconnectionless mode

Sect.4.2 _{The Network Layer Architecture}

The networklayerrequires its own mini architecture because ofthe wide _variability of

subnetworkservices andtechnology. Inordertoallowfor changing_technology and ease of_interfacing,thenetworklayer has been divided intothree sublayers[40,66,68]:

SNICP

-theSubNetwork Independent ConvergenceProtocol SNDCP

-the_{SubNetwork Dependent Convergence}Protocol SNAcP

-theSubNetworkaccessConvergenceProtocol

SNICP

Thisprotocolisbuiltonthemost minimal set of services provided_bya given subnetwork

access protocol. The idea here is to build a single protocol which _{may be} used

universallybyalltheend systems and_{internetworking}units_(gateways)ontheinternet. It is therefore known in advance, that this protocol will not require services of _any

subnetwork,that the subnetworkwill notbe ableto provide. ISO 8348/AD1 (ISO_IP) is an example of anSNICP.

SNDCP

This protocolwilleither providetheOSI network_service, or_{possibly be}usedtoprovide

the service required _by the SNICP. This protocol _{may be} used to enhance a given

subnetwork service and_bringit up tothelevel ofservice required_by the networklayer. Anexample ofthisistheuse of an SNDCPwiththe 1980versionX.25. This version of

X.25 isdeficient inthe _ability to transferinformation regarding ISOnetwork connection

establishment [66]. ISO has defined a _standard, ISO 8878, which describes how one

would implement the SNDCP required to _bring this version of _{X.25 up} to the level required_bytheOSInetworklayerservices.

SNAcP

isnot requiredtoprovidethenetworklayerserviceonitsown. Itisusualthat theSNICP,

or a combination of_SNICP/SNDCP,coupledwiththeSNAcPcometogethertofulfillthe

requirements ofthenetworklayer.

AsisthecasewiththeOSIreference_model, themini architectureforthenetworklayeris

notin itselfa specificationof_any given protocols. Aswill be showninthenext section,

itis sometimes thecasethatasingleprotocol (1984_X.25)canbe usedtorealize allthree

Sect.4.3- General Network Layer

Services

Thenetworklayerofthe ISOreferencemodel is_verycomplex. Thisisprimarily due to

thediverse subnetworksandprotocolswhich reside atthislevel. In general,thenetwork

layer protocols provide a meansfor _transmitting Network Service Data Units _(NSDUs)

across a _subnetwork,or a network ofsubnetworks. _{This strongly implies}the needfora

routing function as well. When the first standards began to emerge forthis _{layer, they}

were connection-oriented. _Later, standards _providing connectionless services were

developed.

The OSI standards for connection-mode operation describe the network layer as

providing themeans of_establishing a_{connection, transferring} NSDUs betweentransport

entities, and _terminating connections. The connectionless-mode operation need _only

providethe meansfor datatransfer. Thenetworklayerprovidesthe transportlayerwith a

transparent view of data transmission across _{potentially heterogeneous} networks.

Routingisthereforeanimportant functionofthenetworklayer.

General CharacteristicsoftheNetworkServices

The Network Service _(NS) provides fortransparent transfer ofdata between NS users.

Specifically,theNSprovides_[32]:

Independence from the_underlying subnetwork technology. This shields theuser

fromthepeculiarities of a givensubnetworkinterfaceorinterfaces.

End-to-end transfer. The network service _{(connection-mode)} provides for

end-to-enddatatransferbetween NS users.

Transparencyoftransferreddata. TheNSmakesno attempttointerpretor restrict

thecontent ofthedata between NSusers.

Service and _quality selection. The NS provides a means _by which the NS user

may agree upon the levelof service _qualityforthe transfer ofdata. _Quality of

reliability.

NS user addressing. The NS provides a means of_{uniquelyidentifying}a given NS

user,using aNetworkService AccessPoint _(NSAP)address.

Theservices andfunctionsofthenetworklayer include:

addressing: The transport layermust _supply a unique _name, so as to _identify the

locationoftheprocess atthedestination host.

expedited data service: this denotes the _ability to deliver data in a prioritized

manner. This means expedited _{data may be delivered}ahead of_previously sent

"normal"

data. The intent here istooffer a service_bywhich_{priority information}

canbypass flowcontrolfacilities at a givenNetwork layerentity.

reset service: This allows the transport _entity to abortthe currentcommunication

processing. This would allow data transmission from the point ofreset, after

synchronizingtheendsystemtransportentities.

routingand relaying: Theprocessof_{sending data from}onesystemtoanother_may

necessitate a multiple_tripthrough oneormoredifferentsubnetworks.

segmentation:Thisinvolvesthe_breakingupofNSDUswhicharetoolargetopass

througha given subnetwork. The offeringofthis serviceimplies the _availability

of a_reassembly service as well. This service providesthe means of_puttingthe

original_{NSDU together, from its} segmented parts.

flowcontrol: This servicehelpstopreventdatafrom_beingexchangedina manner

thatoverwhelmstheend orintermediate systems.

error control: This service gives the transport layer a means for _determining

problems in the network layer entities which _may reside between the end

systems.

typeofservice (TOS): This serviceallows the transportlayerto _specify the type

expected, the networklayer_{may be}enhanced to providetherequired degreeof

Sect.4.4 - OSI

InternetworkingStrategies

Thereare_{basicallytwo strategiesforthe}

internetworkingofOSIsystems[66]:

Hop-by-Hop

InternetApproach

Hop-by-Hop

This approach takes theindividual subnetwork and enhances it to the level required_by

the networklayer service. As an _example, an SNDCP wouldbe used toimplement the features needed _by the network _layer, and make _{up for anything lacking} in the

subnetwork services. The 1984 version ofX.25 is an example ofa subnetwork access

protocol which implements the network layer services without the needfor a SNDCP. The 1980version ofX.25 is an exampleofasubnetworkaccessprotocol which requires

theuse of an SNDCPto_{bring it up}to thelevelrequired_bythenetworklayerservices.

Internet Approach

This approach involves the _building a single _SNICP, which would include all the

enhancements needed_bythesubnetworkwiththemostminimal ofservices.

The _hop-by-hop approachhasthe advantages of_takingadvantageofthe services offered

by a given subnetwork. On theother hand, thedisadvantageofthis is that a different

SNDCPisrequiredfor every different SNAcP. The advantageoftheinternetapproachis

thatoneprotocolcanbe used with all SNAcPs. The disadvantage ofthis approachis that

it assumes _only minimal _{functionality} fromthe _SNAcP, anddoes nottake advantage of

builtinservices as_theyexistinthevarious subnetworks.

OSI Network Connection Types

The networkconnection services offered bythe networklayer services can be grouped

into two categories: connection-oriented and connectionless. The remainder of this

The next section will describe the services offered _byboth flavors ofthe OSI Network

Sect.4.5

-Connection-orientedNetwork Service_(CONS)

This categoryof network connectioncontainsthe_followingprovisions:

aconnection which_maybe used_by thenetwork service users forthe transfer of

data.

a level of_quality across the connection that _may be negotiated _by the network

service providers and users.

theexchange ofNSDUsover a connection.

expedited_data,orthe_abilityto_specifyspecial controlover certainNSDUs.

a reset service which provides the _ability to abort andresume a connection in a controlled manner.

adataconfirmation service.

a network connectionterminationservice.

The CONS approachinvolves thecreation ofalogicalpath, called avirtualcircuit,over

which the NSDUs will be transmitted. Each intermediate system will have a logical connectionto thenextsysteminthepathto theenddestination system. Oncethevirtual

circuitisestablished, allthe NSDUstransferredbetweentheendsystem willtraverse the same path. Itshould benoted thatsince the _{CONS is merely} an_interface, it ispossible that the _underlying subnetwork service is _inherently connectionless. It is not, however,

clear what the advantages wouldbe of _{arbitrarily providing} the CONS in this matter.

This method has several advantages. _First, once avirtual circuitis _established, itis no

longer necessary to considerthe end system address in the NSDUs. An intermediate

system need _{only look in its} tables as to determine where data _{coming in} on virtual

circuitXneedstobesent out on. This simplifiesthe_{routingfunctionalityneeded,} aswell

as the amount of_processingresources needed toroute theNSDUs. _Also, since a single path is used to link the end systems and the intermediate systems, there is no

Flow isundermorecontrol_using CONS. Thisallowstheintermediate systemsto throttle

data accuratelyso thatanintermediateor end systemisnot overwhelmed.

ConnectionlessNetworkCharacteristics

The CLNS had its beginnings in datagram style_networks, such as ARPANET [40]. As

has been shown in previous _chapters, these types of networks _rely on the user (i.e.

transport layeror_above) to ensuredataintegrity. The ISO has issued a standardforthe

specification of aconnectionless-modenetwork service[32].

Connection-orientedNetworkService_(CONS)

The standardforCONS is specifiedin ISO 8348 _[32] andis described in [40,45,66,68].

Thisstandarddefinesthenetwork servicewithregardto:

theprimitive actions and eventsoftheservice

theparameters associatedwitheach primitiveaction andevent.

theinterrelationshipsbetweenvalidsequences oftheevents

Amajor goal ofthisstandardistoprovide a setofservicesthatcanbe appliedto_existing

heterogeneous subnets. This could _potentially allow the_{interconnection necessary for}

global communication.

Background

The network service provides a meansforthe transparent transferofdata _{(NS-user-data)}

between NS users. It is meant toprovide this service in a_way thatdoes not reveal the

techniques used_bythe _underlyingsubnetworkservices. Thisis animportantadvantage,

as these subnetwork services do not _generally possess a common _interface, between

vendors.

There exists a set of parameters which can be used to define the _quality of service

specifyeitherspeedor_{accuracy/reliability}[66]:

Phase Speed

connection_{establishment NCestablishment}

delay

datatransfer throughputtransit_delay

connectiontermination NCrelease_delay

Accuracy/Reliability

NCestablishmentfailure

probability

resilience of transfer

NCtransferfailure

probability

NCreleasefailure

probability

NCEstablishment_Delay

Thisvaluerepresentsthemaximum allowable_delaybetweenan N-Connect.requestand

the_{corresponding N-Connection.confirm.}

NC Establishment Failure_Probability

Thisrepresentstheratio of connection establishment attempts thatfails as a result ofNS

providermalfunction. This includesmisconnection,NSrefusal,and excessivedelay.

Throughput

Thisrepresents theamount ofNSDU transfer that theprovider can_ablysustain. Desired

andmaximumallowable values canbe specified.

Transit_Delay

This value represents the average elapsed time between an N-data.request and the

correspondingN-data.indication. Desiredandmaximumacceptable values are specified.

Residual Error Rate

This equatesto: (incorrect NSDUs +Lost NSDUs+ duplicate _NSDUs) divided_byTotal

TransferFailure_Probability

This valuepertains to_throughput, transit_delay, and residual error rate. This represents

the observed proportion of time that the NS provider fails to provide the minimum

acceptable service.

NCResilience

Theseparameters _specifythe_probabilityof an NS providerinvokedNC release,andNS

providerinvokedreset.

NCRelease_Delay

This represents the maximum acceptable _delay between an NS user invoked N-Disconnect.requestandthesuccessful releaseoftheNCatthepeerNSuser.

NC ReleaseFailure_Probability

This is theratio oftotalNCrelease requests thatresultin failure to the totalnumber of

release requestsissued.

NC Protection

Thisparameter isused to_specify theextent towhich an NSprovider attempts toprevent

unauthorized _tampering, misauthentication, or _monitoring of NSDUs. There are 4

featureswhich_{may be} selected:

confidentiality

modification,deletion,replay,orinsertionofNSDUs NSDUoriginauthentication

NC_Priority

Thisparameter specifies therelative importance of_{gaining/keeping} an NC _connection,

MaximumAcceptableCost

This represents the maximum acceptable cost for an NC. This _may be specified in

absolute orrelativecost units.

Connection-modePrimitives

The_followinglistrepresentstheservice primitives and associatedparameters, fora given

phase ofconnection:

NCEstablishment

N-Connect.request (calledaddress, callingaddress,receipt confirmation selection,

expediteddataselection,QOS parameter_{set,NS-user-data)}

N-Connect.indication (called address, calling address, receipt confirmation

selection,expediteddata_selection,QOSparameter _{set,NS-user-data)}

N-Connect.response _(responding address, receipt confirmation _selection,

expediteddataselection, QOSparameter_{set,NS-user-data)}

N-Connect.confirm_(respondingaddress,receiptconfirmation selection, expedited dataselection,QOS parameter_{set,NS-user-data)}

Data Transfer

N-Data.request (NS-user-data,confirmation_request)

N-Data.indication(NS-user-data,confirmation_request)

N-Data-acknowledge.request₍₎

N-Data-acknowledge.indication₍₎

N-Expedited-data.request_{(NS-user-data)}

N-Expedited-data.indication_{(NS-user-data)}

N-Reset.request(reason)

N-Reset.indication_(originator,reason)

N-Reset.confirm₍₎

NC Release

N-Disconnect.request_{(reason, NS-user-data,respondingaddress)}

N-Disconnect.indication_(originator,reason, NS-user-data,respondingaddress)

The_proceeding sectionswilldescribe each ofthe threephases of a connection. Included

willbeadescriptionoftheservice primitivesusedandtheparameters associatedwiththe

primitives.

NetworkConnection EstablishmentPhase

The NCestablishment service primitives are used to establish an NC between NS users.

The NS users must exist and be known to the NS provider. Simultaneous N-Connect

requests are_{handled separately} and_mayresultin_zero,one,ortwoNCs.

The_followingsectiondetailseach parameterassociatedwithconnectionestablishment.

BackgroundonAddresses

Thereare threenetwork connectionparameterswhichare considered as addresses:

called address

callingaddress

respondingaddress

These are all regarded as network service access points (NSAPs). The NSAP address

parameters accommodate variable length addresses, up to a maximum of 40 decimal

digits. The values contained in these parameters are not _necessarily checked or

authenticated_bytheNSprovider.

Called Address

CallingAddress

Thisparameterconveystheaddress oftheNSAPfromwhichtheNC has beenrequested.

RespondingAddress

Thisparameter conveystheaddress oftheNSAPtowhichtheNC has beenestablished.

ReceiptConfirmation Selection

This parameter selects whetheror notto use or make availablethe receipt confirmation

service. This serviceisnot_mandatoryandthereforeneednotbeprovided_by thelocal,or intermediate,NSprovider(s).

Expedited Data Selection

Thisparameterisusedtoselectthe_{use/availability}oftheexpediteddatatransferservice.

QOS Parameter Set

All QOS parameter setsexchanged _during NC establishment has associated with it the

followingsub-parameters:

targetvalue (value desiredbythe_{calling NS}user)

lowest_quality acceptable(lowestvalueacceptable _bytheNS user)

availablevalue(value NS provideris_willingto_provide)

selectedvalue(valuethecalledNS useragrees_to)

Throughput and transit _delay are _currently the _{only QOS} sub-parameters that are

negotiated _during NC establishment. All other QOS sub-parameters are _currently providedfor inamannernotspecified_bythenetworkservicestandard.

NS-user-data

Thisparameter_actuallystoresthedata_beingexchanged_byNS users. The amount ofdata

normalparameterfortransferofdata. Itsuseis_{mainly for}the transferof set_{up data.}

SequenceofNCPrimitives

The_followingdiagramshowsthesequence of eventsforsuccessfulNCestablishment:

N-Connect Request

N-Connect Confirm

N-Connect Indication

N-Connect Response

NC Establishment Sequence

Note: foranexplanation oftheevent sequence_diagrams,pleaserefertoAppendix A.

Thisconnection establishment_{may fail due}to theNSprovider_beingunableto set_up the

connection. _{It may} also fail due to the refusal of the called NS user to accept the

N-Connect.indication. _Also,theNCestablishment_{may be} aborted_bytheNSprovider,or

either ofthe twoNSusers priorto the_issuingoftheN-Connect.indication.

Network Connection Release Phase

The NC release primitives are usedto release anNC. Therelease can beperformed_by

eitherofthetwoNS users. The NC mayalsobereleased_by theNSprovider. Thecalled

NS user_{may issue} anN-Disconnectas a _wayof_rejecting anN-Connect.indication. The

NS provider_may alsoissue a release if itcannot provide thedesired NC. Oncean NC

releaseisinvokedat oneendpoint,all _{remaining data}and requests forthe other endpoint

may be discardedbytheNS provider.

NC Release Parameters

Originator

Thisparameterindicates thesource oftheNCrelease. Itindicateseither an NS user, NS

provider,orhas thevalue "undefined".

Reason

This parameter conveys information with regard to the cause of the NC release. The

followingvalues are allowedforthisparameter:

Originator=_NSprovider

disconnection

-permanent condition

disconnection

-transientcondition

connect rejection- NSAP

address unknown/permanent condition

connect rejection- NSAP

unreachable/transient condition

connect rejection- NSAP

unreachable/permanent condition

connect rejection- QOS

unavailable/permanent condition

connect rejection- QOS

unavailable/transient condition

connectrejection

-reasonunspecified/permanent condition

connectrejection-reasonunspecified/transient condition

Originator=_NSuser

disconnection

-normalcondition

disconnection - _{abnormal condition}

connectionrejection- _{permanent condition}

connection rejection- _transient

condition

connection rejection- QOS _{unavailable/transient}condition

connectionrejection- QOS_{unavailable/permanentcondition}

Originator=

undefined

reasonparameteralways equals undefinedinthiscase

NS-user-data

Thisparameterallowsforthe transferofdatabetween NSusers. An NS user whichcalls

thereleasefunctionisable tosend0to 128octetsof_{data along}withtheNCrelease.

RespondingAddress

Thisparameterispresentinthecase where rejection ofanNCoccurs_bythecallfromthe

NS user. This parameter contains the NSAP address from which the rejection was

issued.

SequenceofEvents for_ReleasinganNC

The_followingare valid sequencesforthereleaseof an NC:

one NS user_requesting release and the other NS user_receiving an NC release

indication

both NS users_{simultaneously invoke NC}release

theNSproviderinvokes NCrelease,withNCreleaseindications goingtoboth NS

users

both NS user and NS provider release the NC _{by requesting NC} release and

N-Disconnect

Request

N-Disconnect

Indication

NS UserInvokedRelease

N-Disconnect Request

O,

^

Simultaneous NS User Invoked Release

N-Disconnect Indication _O.

^

^j

NS Provider Invoked Release Simultaneous NS User/ProviderRelease NC Release Sequences

An NC connection_{may be}rejected_bytheotherNS user. Inthiscase, NS userAissues

anN-Connect.request andtheother NS user, B, receives an N-Connect.indication. NS

userB doesnot wanttheconnection, andissues an _{N-Disconnect.request,}andNS userA

receivesthe_{corresponding N-Disconnect.indication.}

N-Connect Request N-Disconnect Indication N-Connect Indication N-Disconnect Request

NC Rejection_byDestinationNS User

An NC may alsobereleased_by theNS provider. This case can be illustrated_by an NS

user_issuing anN-Connect.request and_{receiving back} an _{N-Disconnect.indication. The}

rejectedthe connection:

N-Connect Request

N-Disconnect Indication

NC Rejection_by NS Provider

Data Transfer Phase

Thedatatransferphase providesforthe transmissionofdata between NS users. The data

istransferred intheformof_NSDUs, and_{may be}sentineither_direction, simultaneously,

ontheNC.

Data Transfer Primitives

The_followingprimitives existfor providingthedatatransferfunction:

N-Data.request

N-Data.Indication

Data Transfer Parameters

The_followinglistsouteach parameter associated with the datatransferprimitives, and a

descriptionoftheparameter'sfunction.

NS user-data

Thisistheparameterthatallowsthedatatobetransferredbetween NSusers.

Confirmation Request

is desired_bythe sender. TheN-Data-acknowledgeprimitivesprovideamethodforsuch

confirmation. Thisparameter is used_{only for NCs} where negotiation forconfirmation

service wasagreedto_duringtheNCestablishment phase.

SequenceofEventsfor TransferofData

The sequence of primitives in a successful data transfer is shown in the _following

diagram:

N-Data Request

N-Data Indication

Data Transfer Event Sequence

Receipt Confirmation Service

Thereceiptconfirmation serviceisusedinassociation withspecificationof theservicein

theN-Data.request. Each NSDUsentin anN-Data.requestwithreceipt confirmation set

is acknowledged with an N-Data-acknowledge.request. These confirmation of receipt

(COR) will not be combined with previous or future CORs; _they are sent back to the

originating NS user in the same order as the NSDUs are received. In this way, the

originating NS user_maycount acknowledgmentstodeterminewhich NSDUshave been

acknowledged. _Again,thereceipt confirmation serviceis negotiated atNCestablishment phase_bythe twoNS usersandtheNSprovider. It isnot_mandatorythat theNS provider

providethisservice.

The_followingprimitives are suppliedforuseinthereceipt confirmation service:

N-Data-acknowledge.indication

Thereare noparameters associatedwiththeseprimitives.

SequenceofEventsforReceipt Confirmation

The _following diagram illustrates the sequence ofevents forreceipt confirmation ofa

successfulNSDUtransfer:

N-Data Request

N-Data

Acknowledge-Indication

N-Data

^

Indication

~

N-Data Acknowledge

Request

SequenceofEventsfor Receipt Confirmation

An NS user is not allowed to acknowledge an N-Data indication if the confirmation

receiptisnot set. _Further, anNS useris notto send a secondconfirmationreceipton an

N-data.indicationwhich_{has already been}acknowledged.

While in theprocess of _{acknowledging N-Data.indications,} an NS user _mayreceive an

N-reset indication or confirm. Atthis_point, the NS is not allowedto acknowledge _any

N-Data indicationswhichitreceived_upuntilthe time thereset procedure completed.

Expedited Data Transfer Service

The expediteddatatransferfunctions as an alternate methodfor sending dataon an NC.

The same _sequencingrules for NSDUs applytoENSDUs (Expedited NSDUs). The NS

provider guarantees _they will be delivered in the same order as _they were sent. These

ENSDUs havethe special properties ofbeingaccepted_bythe_{receiving NS} userahead of

ENSDUscanbeaccepted_by anNS user whichis no_{longer accepting NS-user-data. The}

use ofthis serviceis negotiated_during the NC establishment phase. It isnot_mandatory fortheNS providertoprovidethisservice.

Expedited Data Transfer Primitives

The_followingprimitivesare usedtorealizetheexpediteddatatransferservice: N-Expedited-data.request

N-Expedite-data.indication

Bothprimitiveshave associatedwiththen a single parameter: NS-user-data. The sizeof

the_{NS-user-data may}rangefrom 1 to32octets.

SequenceofEvents for Expedited Data Transfer

The _following diagram illustrates the sequence of primitives for the expedited data transferfunction:

N-Expedited Data Request

N-Expedited Data Indication

Expedited Data Transfer Sequence

The Reset Service

Thereset service_{may be}usedfortworeasons:

byanNSusertosynchronize use oftheNC

by the NS provider to report loss of data which is unrecoverable within the

Reset_{Service Primitives}

Theresetserviceoffersthe_followingprimitivesto_carryouttheresetfunction:

N-Reset.request

N-Reset.indication

N-Reset.response

N-Reset.confirm

There are twoparameters associated with the reset service: originator and reason. The

originator parameteristhe value usedto indicatethe sourceofthereset. Possiblevalues

are "NS user", "NS provider", or "undefined". The reason parameterhelps to give the

reason forthereset. Iftheoriginator oftheresetisthe NSprovider, then theresetreason

can eitherbe"congestion" or"reason unspecified". WhentheNS userinvokes _reset, the reasonis simply "user synchronization". When theoriginator value is _"undefined", the

reasonvalueisalso "undefined".

EventSequencefor Reset

Therearetwosequences_{whereby NS}userandNS providerexchangeN-Resetprimitives: An NS user issues an N-Reset.request and the NS provider responds with an

N-Resetconfirm

N-Reset Request N-Reset^ Confirm r^, v_> N-Reset Indication N-Reset Response

NS User Invoked Reset

N-Reset Request N-Reset Confirm o.

Ky

o,

^

Simultaneous NS User Invoked Reset

N-Reset Indication N-Reset"* Response o,

^

^j

v^

O

NS Provider Invoked Reset Simultaneous NS User/Provider Reset

Sect.4.6 _{Connectionless}ModeNetwork Service_(CLNS)

This section will detail the services provided _by the network layer for the purpose of

providing a _{connectionless} mode network service. The standard for the CLNS is

specifiedin ISO 8348 Addendum _{1. Itisinterestingin itself}thatthis servicedefinition is

an addendum and not part ofthe original network service specification. The standard

mentions that it became _necessary to define a connectionless mode network service. There are nospecific reasons givenfortheinclusionofthis service,exceptthat_they felt

the connection-oriented service _{unnecessarily limited} thescope ofthereference model. Theinclusion oftheserviceinthenetwork servicedefinitionwill be discussed ingreater detail intheOSI analysis section.

Background

The CLNS doesnot require a connection asdoes theCONS. Therefore, there isnoclear

relationship between units ofdata. Eachunit ofdata may berouted separately, and_may arrive atthe final destination in an orderdifferent from that which was sent. Because

thereisnoconnection _set-up, andtheassociated overheadforsuch_activity, the CLNS is

especiallyattractiveforapplications whichrequireshort-term_{request/reply}transactions.

QualityofNetworkService

Thereis nopeer-to-peer negotiation of_qualityofservice atthe time the transmission of datatakesplace. Thisisbydefinitionoftheconnectionless-modeservice asdescribed in ISO 7498 [25]. A knowledgeofthe_quality of service characteristics are expectedto be knownpriortothetransmission, betweentheNSusers andtheNSprovider. Knowledge

oftheconnectionless-mode characteristicsareavailabletoa_{sending NS}userviaacontrol

or managementfacility. Thefollowingquality of serviceparameters aredefined for the CLNS:

-Transit_Delay

ResidualError_Probability

Priority

Transit_Delay

Thisparameteristheelapsedtime between anN-Unitdata.requestandthe_{corresponding}

N-Unitdata.indication. This value is calculated _only on NSDUs that are _successfully

transferred. Success is determined _by the NSDU _leaving the _{originating NS} user and

arriving at the destination NS user. Transit _delay can change_dynamically andis kept

trackof_bytheNS provider. Thevalueis availableto the NS userand can beevaluated

priorto_sendingtheNSDU.

Protection From Unauthorized Access

This QOS parameter allows the NS user to _specify to the NS provider to choose the

lowestcost methodfortransport. _{It may}also_specifythemaximumallowable cost.

Residual Error_Probability

This parameter denotes the _probability that a single NSDU will be _{lost, duplicated,} or

delivered incorrectly.

Priority

This parameter allows the NS user to _specify the relative _priority of an NSDU. This

priority may have an effect on which NSDUs are to have their _quality of service

degraded,orwhichones are_likelytobe discarded ifresources areinshort supply.

Connectionless-mode Primitives

These functions serve to transfer NSDUs from NSAP to NSAP. All NSDUs are

transferred _{independently} ofeach other. Thereis no maintenance of state information

between a pair ofNSAPs, andtheflowcontrol available exists_only betweentheNSuser

There aretwoprimitives associatedwiththeCLNS:

N-Unitdata.request (source address, destination address, quality of _service,

NS-Userdata)

N-Unitdata.indication (source address, destination address, quality of _service,

NS-User_data)

Inadditionto theservice_primitives,the_followingparameters are associatedwiththem:

Source Address -This parameter specifies the NSAP ofthe _{originating NS} user.

Thisaddressis formatted_identicallyintheOSI CONS.

Destination Address This parameter specifies the NSAP ofthe destination NS

user. Thisaddressisformatted_identicallyintheOSI CONS.

Quality of Service this value specifies the values for each of the QOS

sub-parameters. Fortherequest_primitive, thevalues can be anythingwhich is

defined. Forthe indication primitive, these values _{may be different}than those

listedintherequest primitive.

NS-Userdata This parameterallows the transferdata between NS users. _Any

number of_octets,from 0to_{64,512,may be}sent.

EventSequencefor Data Transfer

The _following diagram shows the sequence of events for N Unitdata.request and

N-Unitdata.indication:

N-UnitData. Request

N-Unit Data Indication

Sect.4.7- Protocolsfor

ProvidingtheOSINetwork Service

The next 2 sections will describe the protocols for providing the OSI network layer

services. The first section will describe the protocols whichhave been standardizedby the ISO for _providing the connection-mode network service. The second section will describe the protocol _{for providing} the connectionless-mode network service. From a

surveyofthestandards documents _[33,36,37] whichdescribeprotocolsfor providing the

network_service,thereisan attempttodescribetheuse of_existing networkanddata link protocols. These are meant toserve as guidelines, and in no_way mandate the use ofa