RFC2544 测试规范文档格式.docx
《RFC2544 测试规范文档格式.docx》由会员分享,可在线阅读,更多相关《RFC2544 测试规范文档格式.docx(40页珍藏版)》请在冰豆网上搜索。
![RFC2544 测试规范文档格式.docx](https://file1.bdocx.com/fileroot1/2023-1/22/d3ac9b1c-3706-4e1f-bcd6-63702cff2e51/d3ac9b1c-3706-4e1f-bcd6-63702cff2e511.gif)
RFCreplacesandobsoletesRFC1944.
Abstract
Thisdocumentdiscussesanddefinesanumberofteststhatmaybe
usedtodescribetheperformancecharacteristicsofanetwork
interconnectingdevice.Inadditiontodefiningtheteststhis
documentalsodescribesspecificformatsforreportingtheresultsof
thetests.AppendixAliststhetestsandconditionsthatwebelieve
shouldbeincludedforspecificcasesandgivesadditional
informationabouttestingpractices.AppendixBisareference
listingofmaximumframeratestobeusedwithspecificframesizes
onvariousmediaandAppendixCgivessomeexamplesofframeformats
tobeusedintesting.
1.Introduction
Vendorsoftenengagein"
specsmanship"
inanattempttogivetheir
productsabetterpositioninthemarketplace.Thisofteninvolves
"
smoke&
mirrors"
toconfusethepotentialusersoftheproducts.
Bradner&
McQuaidInformational[Page1]
RFC2544BenchmarkingMethodologyMarch1999
Thisdocumentdefinesaspecificsetofteststhatvendorscanuseto
measureandreporttheperformancecharacteristicsofnetwork
devices.Theresultsofthesetestswillprovidetheusercomparable
datafromdifferentvendorswithwhichtoevaluatethesedevices.
Apreviousdocument,"
BenchmarkingTerminologyforNetwork
InterconnectDevices"
(RFC1242),definedmanyofthetermsthatare
usedinthisdocument.Theterminologydocumentshouldbeconsulted
beforeattemptingtomakeuseofthisdocument.
2.Realworld
Inproducingthisdocumenttheauthorsattemptedtokeepinmindthe
requirementthatapparatustoperformthedescribedtestsmust
actuallybebuilt.Wedonotknowof"
offtheshelf"
equipment
availabletoimplementallofthetestsbutitisouropinionthat
suchequipmentcanbeconstructed.
3.Teststoberun
Thereareanumberoftestsdescribedinthisdocument.Notallof
thetestsapplytoalltypesofdevicesundertest(DUTs).Vendors
shouldperformalloftheteststhatcanbesupportedbyaspecific
typeofproduct.Theauthorsunderstandthatitwilltakea
considerableperiodoftimetoperformalloftherecommendedtests
nderalloftherecommendedconditions.Webelievethattheresults
areworththeeffort.AppendixAlistssomeofthetestsand
conditionsthatwebelieveshouldbeincludedforspecificcases.
4.Evaluatingtheresults
Performingalloftherecommendedtestswillresultinagreatdeal
ofdata.Muchofthisdatawillnotapplytotheevaluationofthe
devicesundereachcircumstance.Forexample,therateatwhicha
routerforwardsIPXframeswillbeoflittleuseinselectinga
routerforanenvironmentthatdoesnot(andwillnot)supportthat
protocol.Evaluatingeventhatdatawhichisrelevanttoa
particularnetworkinstallationwillrequireexperiencewhichmaynot
bereadilyavailable.Furthermore,selectionoftheteststoberun
andevaluationofthetestdatamustbedonewithanunderstandingof
generallyacceptedtestingpracticesregardingrepeatability,
varianceandstatisticalsignificanceofsmallnumbersoftrials.
McQuaidInformational[Page2]
5.Requirements
Inthisdocument,thewordsthatareusedtodefinethesignificance
ofeachparticularrequirementarecapitalized.Thesewordsare:
*"
MUST"
Thisword,orthewords"
REQUIRED"
and"
SHALL"
meanthat
theitemisanabsoluterequirementofthespecification.
SHOULD"
Thiswordortheadjective"
RECOMMENDED"
meansthat
theremayexistvalidreasonsinparticularcircumstancesto
ignorethisitem,butthefullimplicationsshouldbe
understoodandthecasecarefullyweighedbeforechoosinga
differentcourse.
MAY"
OPTIONAL"
meansthatthis
itemistrulyoptional.Onevendormaychoosetoincludethe
itembecauseaparticularmarketplacerequiresitorbecauseit
enhancestheproduct,forexample;
anothervendormayomitthe
sameitem.
Animplementationisnotcompliantifitfailstosatisfyoneormore
oftheMUSTrequirementsfortheprotocolsitimplements.An
implementationthatsatisfiesalltheMUSTandalltheSHOULD
requirementsforitsprotocolsissaidtobe"
unconditionally
compliant"
;
onethatsatisfiesalltheMUSTrequirementsbutnotall
theSHOULDrequirementsforitsprotocolsissaidtobe
conditionallycompliant"
.
6.Testsetup
Theidealwaytoimplementthisseriesoftestsistouseatester
withbothtransmittingandreceivingports.Connectionsaremade
fromthesendingportsofthetestertothereceivingportsofthe
DUTandfromthesendingportsoftheDUTbacktothetester.(see
Figure1)Sincethetesterbothsendsthetesttrafficandreceives
itback,afterthetraffichasbeenforwardedbuttheDUT,thetester
caneasilydetermineifallofthetransmittedpacketswerereceived
andverifythatthecorrectpacketswerereceived.Thesame
functionalitycanbeobtainedwithseparatetransmittingand
receivingdevices(seeFigure2)butunlesstheyareremotely
controlledbysomecomputerinawaythatsimulatesthesingle
tester,thelaborrequiredtoaccuratelyperformsomeofthetests
(particularlythethroughputtest)canbeprohibitive.
McQuaidInformational[Page3]
+------------+
||
+------------|tester|<
-------------+
||||
|+------------+|
+----------->
|DUT|--------------+
Figure1
+--------++------------++----------+
||||||
|sender|-------->
|DUT|--------->
|receiver|
Figure2
6.1Testsetupformultiplemediatypes
TwodifferentsetupscouldbeusedtotestaDUTwhichisusedin
real-worldnetworkstoconnectnetworksofdifferingmediatype,
localEthernettoabackboneFDDIringforexample.Thetestercould
supportbothmediatypesinwhichcasethesetupshowninFigure1
wouldbeused.
TwoidenticalDUTsareusedintheothertestsetup.(seeFigure3)
Inmanycasesthissetupmaymoreaccuratelysimulatethereal
world.Forexample,connectingtwoLANstogetherwithaWANlinkor
highspeedbackbone.Thissetupwouldnotbeasgoodatsimulating
asystemwhereclientsonaEthernetLANwereinteractingwitha
serveronanFDDIbackbone.
+-----------+
+---------------------|tester|<
---------------------+
|+-----------+|
|+----------++----------+|
+------->
|DUT1|-------------->
|DUT2|---------+
+----------++----------+
Figure3
McQuaidInformational[Page4]
7.DUTsetup
Beforestartingtoperformthetests,theDUTtobetestedMUSTbe
configuredfollowingtheinstructionsprovidedtotheuser.
Specifically,itisexpectedthatallofthesupportedprotocolswill
beconfiguredandenabledduringthissetup(SeeAppendixA).Itis
expectedthatallofthetestswillberunwithoutchangingthe
configurationorsetupoftheDUTinanywayotherthanthatrequired
todothespecifictest.Forexample,itisnotacceptabletochange
thesizeofframehandlingbuffersbetweentestsofframehandling
ratesortodisableallbutonetransportprotocolwhentestingthe
throughputofthatprotocol.Itisnecessarytomodifythe
configurationwhenstartingatesttodeterminetheeffectoffilters
onthroughput,buttheonlychangeMUSTbetoenablethespecific
filter.TheDUTsetupSHOULDincludethenormallyrecommended
routingupdateintervalsandkeepalivefrequency.Thespecific
versionofthesoftwareandtheexactDUTconfiguration,including
whatfunctionsaredisabled,usedduringthetestsMUSTbeincluded
aspartofthereportoftheresults.
8.Frameformats
TheformatsofthetestframestouseforTCP/IPoverEthernetare
showninAppendixC:
TestFrameFormats.Theseexactframeformats
SHOULDbeusedinthetestsdescribedinthisdocumentforthis
protocol/mediacombinationandthattheseframeswillbeusedasa
templatefortestingotherprotocol/mediacombinations.Thespecific
formatsthatareusedtodefinethetestframesforaparticulartest
seriesMUSTbeincludedinthereportoftheresults.
9.Framesizes
AllofthedescribedtestsSHOULDbeperformedatanumberofframe
sizes.Specifically,thesizesSHOULDincludethemaximumandminimum
legitimatesizesfortheprotocolundertestonthemediaundertest
andenoughsizesinbetweentobeabletogetafullcharacterization
oftheDUTperformance.Exceptwherenoted,atleastfiveframe
sizesSHOULDbetestedforeachtestcondition.
TheoreticallytheminimumsizeUDPEchorequestframewouldconsist
ofanIPheader(minimumlength20octets),aUDPheader(8octets)
andwhateverMAClevelheaderisrequiredbythemediainuse.The
theoreticalmaximumframesizeisdeterminedbythesizeofthe