路由選擇翻譯

Chapter13.RoutingChoices第13章路由選擇Intheearlydaysofnetworking,hugeprojectswereundertakentodeterminethebest在早期的網(wǎng)絡(luò)、大型項目進行確定最佳routingprotocol一forsomedefinitionof“best.”Forinstance,oneprojectcommissioned路由協(xié)議的一些“最好的定義?！袄纾粋€項目委托bytheUSAirForceattemptedtomeasurethespeedofconvergenceforeachofthe由美國空軍試圖衡量收斂的速度primaryroutingprotocols(OSPF,IS-IS,andEIGRP),howmuchbandwidtheachprotocol主要路由協(xié)議(OSPF是什么,EIGRP)，多少帶寬每個協(xié)議usedundernormaloperation,howmuchbandwidthwasrequiredwhentheprotocolwas在正常的操作下，使用協(xié)議時需要多少帶寬converging(givenaspecificnetworktopologyandbreakagescenarios),howmany融合(給定一個特定的網(wǎng)絡(luò)拓撲和破損的場景),有多少networkengineerskneweachoftheprotocolsunderinvestigation,andanumberofother網(wǎng)絡(luò)工程師知道每個協(xié)議接受調(diào)查，和其他一些factors.Ultimately,theprojectwasdesignedtochurnallthesefactorstogetherand因素。最終，該項目旨在生產(chǎn)和所有這些因素在一起produceasingleanswer.產(chǎn)生一個答案。Underlyingallthesecalculationswasalwayssomesetofassumptions一assumptions基本所有這些計算總是assumptions-assumptions的一些設(shè)置rarelyeverbroughtoutintothelightofdayandclearlyexamined,muchlessfullyjustified.很少了天日，明確檢查，更充分的理由。Theseassumptionsare,inreality,theinterestingpieceofthe“whichroutingprotocol”這些假設(shè)是，在現(xiàn)實中，“哪個路由協(xié)議”的有趣puzzle.Hereyouwillfindtherealintersectionbetweenthebusinessandtechnology謎題。在這里你會發(fā)現(xiàn)真正的業(yè)務(wù)和技術(shù)之間的十字路口choices.選擇。Thefirstpartofthischapterwilldealwiththe“whichroutingprotocolisbest”question,本章的第一部分將處理”，路由協(xié)議是最好的”問題，butwe'llbeconcernedwithexposingthequestionsnetworkarchitectsneedtoaskin但我們會關(guān)心揭露網(wǎng)絡(luò)架構(gòu)師需要問的問題ordertoanswerthatquestion,ratherthanprovidingaonesizefitsallsolution.為了回答這個問題，而不是提供一個一刀切的解決方案。Beyondchoosingaroutingprotocol,therearequestionsaroundbestpracticesin除了選擇路由協(xié)議，有問題的最佳實踐deployingrouting;thisiswherethesecondpartofthischapterconcentrates.Forinstance,部署路由，這是本章的第二部分集中的地方。例如，howshouldfloodingdomainboundariesbesetupinalinkstateprotocol?Arethere洪水域邊界應(yīng)如何建立鏈路狀態(tài)協(xié)議?有trade-offsindifferentplansordesigns?權(quán)衡不同的計劃或設(shè)計嗎？Finally,we'11consideradifferentsetofquestionsaboutroutingprotocoldeployment,最后,我們將考慮一組不同的路由協(xié)議部署問題specificallyrelatedtothecurrent“BGPeverywhere"trendsocommoninlargescale專門針對當(dāng)前“邊界網(wǎng)關(guān)協(xié)議無處不在”的趨勢在大規(guī)模如此常見networksdesigns.Isend-to-endBGPreallythebestanswertoresolvetherouting網(wǎng)絡(luò)設(shè)計。端到端邊界網(wǎng)關(guān)協(xié)議是最好的答案來解決路由problemsmodernnetworksface?Whatproblemsdonetworksdesignedwithend-to-end現(xiàn)代網(wǎng)絡(luò)面臨的問題?什么問題做網(wǎng)絡(luò)設(shè)計的端到端BGPface?邊界網(wǎng)關(guān)協(xié)議的臉嗎？WhichRoutingProtocol?路由協(xié)議？Whatarethetraditionalquestionssurroundingthechoiceofaroutingprotocol?Most有哪些傳統(tǒng)問題周圍的一種路由協(xié)議的選擇?大多數(shù)networkoperatorswilljumptotwoquestions:網(wǎng)絡(luò)運營商將跳轉(zhuǎn)到兩個問題：Howfastdoesitconverge??收斂的速度有多快？258258年Isitproprietary??它是專有的嗎？Thesetwoquestionsquicklybecomethecenterofvirtuallyanydiscussiononchoosinga這兩個問題迅速成為任何討論選擇的中心routingprotocol.Butaretheytherightquestions?Thereisanothersetofquestionsevery路由協(xié)議。但他們是正確的問題嗎?有另一組的每一個問題networkdesignershouldconsider—questionsthatbringtherelationshipbetweenthe網(wǎng)絡(luò)設(shè)計師應(yīng)該考慮質(zhì)疑,帶之間的關(guān)系businessandtechnologymoreintoview.Thesequestionsincludethefollowing:業(yè)務(wù)和技術(shù)更多的視圖。這些問題包括以下幾點：Whichprotocoliseasiertoconfigure??協(xié)議更容易配置？Whichprotocoliseasiertotroubleshoot??協(xié)議更容易解決？Whichprotocoldegradesinawaythatprovidesthebestsupportforthebusinessworks?這協(xié)議降解的方式提供了最佳的對業(yè)務(wù)工作的支持duringfailures?在失敗？Whichprotocolworksbestonthetopologythebusinessusuallybuilds??協(xié)議效果最好的業(yè)務(wù)通常構(gòu)建拓撲？Let'sconsidereachofthesequestionsinturn.讓我們依次考慮這些問題。HowFastDoestheRoutingProtocolConverge?路由協(xié)議收斂有多快？Convergencespeed,althoughoftendiscussed,debated,anddiscovered,isarather收斂速度,雖然經(jīng)常討論、辯論和發(fā)現(xiàn)，是一個相當(dāng)slipperyconcepttodefine.Wemustbeginwiththeratherbenignsoundingquestion:難以把握的概念來定義。我們必須開始而良性測深的問題：Whatdoesitmeanforthenetworktobeconverged?Whentestingforconvergence,是什么意思的網(wǎng)絡(luò)聚合嗎?為收斂測試時，engineersoftensetupsmall,simpletestbedsthatmakeforeasycontroland工程師通常設(shè)置小，容易控制和簡單的測試床measurement.Theythenusevariousmethods(suchasthoseoutlinedinRFC4061,4062,測量。然后他們利用各種方法(如RFC4061中概述,4062年，and4063),todeterminewhentheroutingprotocolhasfinishedcalculatingthebestpath和4063年)，以確定當(dāng)路由協(xié)議已經(jīng)完成了計算最佳路徑toeachpossibledestinationwithinthe(simple)network.Ortheengineermightusea內(nèi)的每個可能的目的地(簡單的)網(wǎng)絡(luò)。或工程師可以使用trafficgeneratortodeterminehowmanypacketsaredroppedinthe(simple)network流量發(fā)生器來確定有多少(簡單的)網(wǎng)絡(luò)數(shù)據(jù)包被丟棄duringaconvergenceevent.在一個收斂的事件。Whileallthismeasurementisfine,itdoesn'treallyanswerreal-worldquestions.Asa雖然所有這些測量是好的,但并沒有真正回答現(xiàn)實問題。作為一個counterexample,let'sconsidertheglobalInternet.TheglobalInternetismadeupoftens反例，讓我們考慮全球互聯(lián)網(wǎng)。全球互聯(lián)網(wǎng)是由數(shù)以萬計ofthousands(ifnothundredsofthousands)ofautonomoussystems.Theseautonomous的(如果不是成千上萬)自治系統(tǒng)。這些自治systemsareinterconnectedbysomewherebetween250,000and1millioneBGPspeakers.系統(tǒng)是相互聯(lián)系的，在250000年和100萬年之間eBGP揚聲器。ThenumberofBGPspeakerswithintheglobalInternet(whichwouldincludeiBGP邊界網(wǎng)關(guān)協(xié)議揚聲器的數(shù)量在全球互聯(lián)網(wǎng)(包括iBGPspeakers,routereflectors,andotherBGPspeakingdevices)easilynumbersoveramillion,揚聲器、路由反射器和其他邊界網(wǎng)關(guān)協(xié)議容易說設(shè)備)數(shù)量超過一百萬，andthetotalnumberofrouters—addingroutersthatarenotBGPspeakersatall—isat和的總數(shù)routers-adding路由器沒有邊界網(wǎng)關(guān)協(xié)議揚聲器在其他leastdoublethetotalnumberofBGPspeakers.邊界網(wǎng)關(guān)協(xié)議揚聲器總數(shù)的兩倍。259259年Thisscale,combinedwiththeconstantchurnofmodifiedpolicies,changingconnectivity,這種規(guī)模，加上不斷攪動修改政策，改變連接，andsimplefailures,producesanetworkthatneverconverges.DoesthismeanBGPisa和簡單的故障，產(chǎn)生的網(wǎng)絡(luò)不會收斂。邊界網(wǎng)關(guān)協(xié)議是一個,這意味著什么“bad”protocolfromaconvergencespeedperspective?No,itsimplymeansdesignersneed“壞”協(xié)議從收斂速度的角度?不，這僅僅意味著設(shè)計師需要toberealisticaboutnetworkconvergence.Togetahandleontherealworld,youneedto以現(xiàn)實的態(tài)度對待網(wǎng)絡(luò)收斂。處理在現(xiàn)實世界中，您需要turnthequestionaround.扭轉(zhuǎn)問題。Thefirstquestionyouneedtoaskis:Howfastdoyouneedthenetworktoconverge?你要問的第一個問題是:你需要網(wǎng)絡(luò)收斂速度？Intherealworld,therewon'tbeoneanswertothisquestion—therewillbeoneanswerfor在現(xiàn)實世界中，不會有一個回答這個問題一個答案everyapplication,everybusinessrequirement,andeveryareaofthenetwork.With每一個應(yīng)用程序中，業(yè)務(wù)需求，每個區(qū)域的網(wǎng)絡(luò)。與carefulplanning,youcanbuildtiersofconvergencerequirements.Thefastest仔細的規(guī)劃，您可以構(gòu)建層收斂要求。最快的convergencetimesshouldbecontainedinasmallsliceofthenetwork（seeChapter14,收斂時間應(yīng)該包含在網(wǎng)絡(luò)的一小部分（見第14章，“DefiningComplexity,”foramorecompletediscussionofthetrade-offbetweenspeedand“定義的復(fù)雜性，為一個更完整的討論速度和之間的權(quán)衡stability）.穩(wěn)定）。Thesecondquestionyouneedtoaskis:Howmuchareyouwillingtopayforfaster第二，你需要問的問題是:你愿意付多少更快convergence?收斂嗎？Givenabaselineofredundantlinksandequipment,fastconvergencerevertstosoftware給定一個基線的冗余鏈接和設(shè)備，快速收斂恢復(fù)軟件andconfiguration.It'softeneasytoseetheadditionalsoftwareandconfigurationas和配置。往往容易看到額外的軟件和配置“free,”intermsofinvestment,whichskewsthetrade-offbetweenfastconvergenceand“自由”，在投資方面，這扭曲了快速收斂和之間的權(quán)衡fasterconvergence.Rememberthedesign,configuration,andequipmentinvestmentin更快的收斂。記住設(shè)計、配置和設(shè)備投資fasterconvergencecanbemuchhigherthanitappearsinmanyareas.更快的收斂可以遠遠高于它出現(xiàn)在許多領(lǐng)域。Forinstance,fasterconvergingnetworksmaybemorebrittle—althoughminorfailures例如，更快的融合網(wǎng)絡(luò)可能更brittle-although小故障maybehandledmorequickly,majorfailuresmaynotbehandledatall.Modifyingthe可以更快的處理，主要的失敗可能不處理。修改topologyinanetworkdesignedaroundfastconvergenceisn'tassimpleasaddinganew拓撲結(jié)構(gòu)在網(wǎng)絡(luò)收斂速度快，設(shè)計并不簡單,添加一個新的linkhereorthere;realanalysisworkmustbedonetoensurethattheoverallconvergence這里或者那里聯(lián)系，真正的分析工作必須完成，以確保整體收斂性propertiesofthenetworkremainstableorimprove.Finally,fastconvergence網(wǎng)絡(luò)的屬性保持穩(wěn)定或改善。最后，快速收斂configurationsaren'toftenwellunderstoodbysupportpersonnel,evenatthenetwork配置通常不會很好地理解支持人員，甚至在網(wǎng)絡(luò)vendor;thiscouldmakeforlonganddifficulttroubleshootingsessionsoverthesmallest供應(yīng)商，這可能使漫長而又艱難的最小故障診斷會話ofproblems.的問題。Whenalltherightquestionshavebeenasked—andanswered—wecanreturntothe當(dāng)所有正確的問題已經(jīng)要求和answered-we可以返回questionathand:Whichprotocolconvergesmorequickly?手頭的問題:哪個協(xié)議收斂更快？EIGRPcanconvergearoundsinglelinkornodefailures,giventhecorrectdesign,inwellEIGRP能集中在單一鏈路或節(jié)點故障，給出正確的設(shè)計，在underasecond.ThecrucialpointtoconsiderinEIGRPfastconvergenceisthedepthof在第二個。EIGRP快速收斂的關(guān)鍵考慮的深度thequeryprocess,orratherhowmanyroutersmustprocessaquerybeforethenetwork查詢過程,或者說有多少路由器網(wǎng)絡(luò)之前必須處理一個查詢canconverge.Asageneralrule,EIGRPnormallyaddsabout200millisecondstothe可以收斂。作為一般規(guī)則,EIGRP通常增加了約200毫秒260260年convergencetimeforeachhopthatmustprocessaquery,sotoconvergeinlessthanone收斂時間必須處理一個查詢每一妝h所以在不到一收斂second,youneedtokeepthequeryscopedowninthetwohoprange.其次，你需要查詢范圍在兩跳范圍內(nèi)。TheupsidetoEIGRPfastconvergenceisthereisalmostnoconfigurationrequiredtoEIGRP快速收斂的好處是幾乎沒有配置要求produceafastconvergingnetwork.SolongastheEIGRPprocesscandetectthefailedlink產(chǎn)生一個快速融合網(wǎng)絡(luò)。只要EIGRP過程可以檢測失敗的鏈接quickly,andthenetworkisdesignedcorrectly,EIGRPwilldoitspart.Thedownsideisthe很快，和網(wǎng)絡(luò)設(shè)計正確,EIGRP都會起到一定的作用。不好的一面是networkmustbedesignedcorrectly,whichmeansalotofthoughtneedstogointowhere網(wǎng)絡(luò)必須正確設(shè)計，這意味著大量的思想需要進入的地方andhowtoplacelinkswhenthenetworkisbeingchangedforanyreason.和如何將鏈接時，網(wǎng)絡(luò)正在改變?nèi)魏卫碛?。OSPFcanconvergearoundsinglelinkornodefailures,giventhecorrectdesignandOSPF可以集中在單一鏈路或節(jié)點故障，給出正確的設(shè)計和configuration,inwellunderasecond,aswell.Thereareseveralkeypointstoremember配置，在1秒內(nèi)。有幾個要點記住whendesigningaroundOSPFconvergence,includingtheplacementoffloodingdomain當(dāng)設(shè)計在OSPF收斂，包括洪水域的位置boundaries,thespeedatwhichLinkStateAdvertisements(LSA)aregenerated,the邊界,鏈路狀態(tài)廣告的速度(LSA)生成的frequencyandtimingofShortestPathFirst(SPF)calculations,andthepathandtimingof頻率和時間最短路徑優(yōu)先(SPF)的計算,和路徑和時間floodingthroughthenetwork.洪水通過網(wǎng)絡(luò)。TheupsidetoOSPFfastconvergenceisthatit'sbetterunderstoodthanEIGRPfastOSPF快速收斂的好處是，它的更好的理解比EIGRP快convergence,andalsothatitconvergesquicklyonawidervarietyofnetworktopologies收斂，也很快就收斂在一個廣泛的網(wǎng)絡(luò)拓撲（atpresstime,therearetechniquesthatwouldallowEIGRPtoconvergejustasquicklyon（截至記者發(fā)稿時，有技術(shù),將允許EIGRP收斂很快thesamesetoftopologies,buttheyhavenotbeenimplemented）.Thedownsideisthat同一組拓撲，但他們沒有實現(xiàn)）。不利的一面是，OSPFfastconvergenceisadelicatedancebetweenconfigurationanddesign.ThedesignOSPF快速收斂之間微妙的舞蹈配置和設(shè)計。設(shè)計gainsalittlemorelatitudeatthecostofextraconfigurationandextradatabase獲得更多的緯度在額外的配置和數(shù)據(jù)庫的成本information,bothofwhichcanmakeconfiguringandtroubleshootingfastconvergence信息，這兩個可以配置和故障診斷快速收斂moredifficult.更加困難。IS-IScanconvergearoundsinglelinkornodefailures,giventhecorrectdesignand是什么可以集中在單一鏈路或節(jié)點故障，給出正確的設(shè)計和configuration,inwellunderasecond,justasOSPFandEIGRPcan.Thekeypointsto配置，在1秒內(nèi)，正如OSPF和EIGRP能。的要點remember（upsidesanddownsides）aboutfastconvergenceinIS-ISaresimilarenoughto記?。ㄕ婧拓撁嬗绊懀焖偈諗渴鞘裁醋阋允窍嗨频腛SPFthattheydon'tneedtoberepeated.ThemaindifferencetoconsideristhatIS-ISOSPF,他們不需要重復(fù)?？紤]的主要區(qū)別是什么hasalittlemoredeploymentlatitudeonvariousnetworktopologies,whichcanmakefast有更多的部署緯度不同的網(wǎng)絡(luò)拓撲，從而使快convergencedesigneithereasierormoredifficult.融合設(shè)計更容易或更困難。IstheRoutingProtocolProprietary?路由協(xié)議專有的嗎？Anotherkeyquestionmostbusinessesareconcernedaboutwhenchoosingarouting另一個關(guān)鍵問題，大多數(shù)企業(yè)都關(guān)心在選擇路由protocoliswhethertheprotocolisproprietary;however,whatdoestheconceptofa協(xié)議是協(xié)議是私有的;然而,的概念是什么proprietaryprotocolreallymean?Generallyspeaking,ifthedesignoftheprotocolis專有協(xié)議的真正意思是什么呢?一般來說，如果協(xié)議的設(shè)計supportedandmanagedbysomeformofstandardsbody,andtheprotocolisavailableon由某種形式的標(biāo)準(zhǔn)組織，支持和管理和協(xié)議261261年morethanonevendor'sequipment,itisconsideredanopen,ratherthanproprietary,不止一個供應(yīng)商的設(shè)備，它被認為是一個開放的、而不是專有的，protocol.協(xié)議。Definingwhataproprietaryprotocolis,however,isjustasslipperyinpracticeasdefining定義什么是私有協(xié)議，然而，一樣滑在實踐中定義fastconvergence.快速收斂。Tounderstandwhy,wemustbeginwithabetterunderstandingofhowstandardsbodies要理解為什么，我們必須首先了解如何標(biāo)準(zhǔn)組織work,andwhattheyproposetostandardize.TheIETFisthestandardsbodymost工作，他們所提出的標(biāo)準(zhǔn)化。IETF標(biāo)準(zhǔn)的身體engineersassociatewiththedevelopmentandmaintenanceofroutingprotocols—butfor工程師與路由protocols-but的開發(fā)和維護alltheprestigeandeffortoftheengineerswhoworkhardabuildingasustainableand所有工程師的威望和努力構(gòu)建一個可持續(xù)和努力工作solidsetofprotocolsonwhichtobuildnetworks,humanrealitiesstillslipin.固體的協(xié)議設(shè)置為構(gòu)建網(wǎng)絡(luò)，人類仍在現(xiàn)實。Fundingcanoftendrivetheadoptionofaprotocol,ratherthantechnical資金通?？梢则?qū)動采用一個協(xié)議，而不是技術(shù)merit—particularlywhereasingleplayerwithsomestakeintheprocessfundsthemerit-particularly,單個球員和一些流程基金的股份developmentandimplementationofonespecificsolution,makingitlooklikethe開發(fā)和實施一個特定的解決方案,使它看起來像adoptionofthatsolutionisallbutinevitable.Personalitycan,likewise,haveamajor采用的解決方案是不可避免的。個性，同樣，有一個專業(yè)impactontheflowofthestandardizationprocess;engineershaveegosjustliketherestof影響流動的標(biāo)準(zhǔn)化過程，工程師們自我就像其他的一樣humanity,andoftenclingtotheirworkfarpastwhentheworkitselfhasceasedtobe人類,經(jīng)常堅持他們的工作遠過去當(dāng)工作本身已經(jīng)停止useful.Competitivepressurescanplayarole,aswell.Difficult-to-readstandardswitha有用的。競爭壓力也起了一定作用。不容易的標(biāo)準(zhǔn)largeamountofimplementationleewayoftenresultfromseveralcompetitorssquaringoff大量的實現(xiàn)空間往往源于幾個競爭對手展開廝殺anddecidingtobuildastandardthatwillallowbothofthemtoclaimstandardsbacking并決定建立一個標(biāo)準(zhǔn)，讓他們兩人聲稱支持標(biāo)準(zhǔn)whileworkingonextensionsandmodificationsthatpreventrealinteroperability.工作時防止真正的互操作性的擴展和修改。Althoughnoneoftheseproblemsareuniquetotheengineeringcommunity,and盡管這些問題是獨一無二的工程社區(qū)，和standardsbodiesoftendowellatrejectingtheeffectsofthesehumanrealities,theystill標(biāo)準(zhǔn)組織經(jīng)常在拒絕這些人類現(xiàn)實的影響，他們?nèi)匀籭mplythatsayingsomethingis“standardsbased”isoftennotenough.Interoperable暗示說“基于標(biāo)準(zhǔn)”往往是不夠的?？苫ゲ僮鞯膒rotocolsmayoftencomeoutofastandardsprocess,butstandardsprocessesdon't協(xié)議可能經(jīng)常出來一個標(biāo)準(zhǔn)的過程，但是不標(biāo)準(zhǔn)流程alwaysresultininteroperableprotocols.Thereisalonglistofprotocolsandideasthat總是導(dǎo)致互操作協(xié)議。有一長串的協(xié)議和想法haveoriginatedinstandardsbodies,orthathavebeenstandardizedinsomeway,but起源于標(biāo)準(zhǔn)組織，或以某種方式被標(biāo)準(zhǔn)化，但havenotsurvivedinthewild.沒有在野外幸存下來。Theobvioussolutiontothisistonarrowthescopeofproprietarysoitincludesnotonly顯而易見的解決方案是狹窄的專有所以它不僅包括的范圍protocolsthathavebeenstandardized,butalsothosethathavebeenimplementedby協(xié)議已經(jīng)標(biāo)準(zhǔn)化，而且那些已經(jīng)實現(xiàn)morethanonevendor.Again,however,thisdefinitionisfraughtwithproblems.Ifwe're不止一個供應(yīng)商。再一次，然而，這一定義是問題重重。如果我們discussingaroutingprotocol,isitsimplementationbyaroutervendorandafirewall討論一個路由協(xié)議，由路由器供應(yīng)商和其實現(xiàn)防火墻vendorenoughtoconsidertheprotocol“open”?供應(yīng)商足夠的考慮協(xié)議“開放”嗎？Thebestsolutionwhendealingwithproprietaryprotocolsistoturnthequestionaround.最好的解決方案在處理專用協(xié)議是扭轉(zhuǎn)問題。Let'sbreaktheshorthandoutintoactualbusinessrequirements.讓我們打破速記到實際的業(yè)務(wù)需求。262262年IsthisprotocolimplementedontheplatformsIwouldlikelyuseinmynetwork??這是協(xié)議在平臺上實現(xiàn)我可能使用網(wǎng)絡(luò)嗎？Insteadofaskingwhethertheprotocolisastandardornot,askwhethertheprotocolis問是否協(xié)議是一個標(biāo)準(zhǔn)的，問的協(xié)議implementedoneachoftheplatformsyou'relikelytoactuallyuse.每個平臺上實現(xiàn)你可能實際使用。Howinteroperablearethevariousimplementationsofthisprotocol?Evenasingle?如何互操作的各種實現(xiàn)這個協(xié)議?甚至一個protocolimplementedmultipletimesbythesamevendorcanhaveinteroperability協(xié)議實現(xiàn)多次同樣的供應(yīng)商可以互操作性problems,eitherbecausethestandardislooseenoughtoallownoninteroperableoptions,問題,因為標(biāo)準(zhǔn)寬松，足以讓互操作的選項，orbecauseoneormoreimplementationsjustaren'twellformed.還是因為一個或多個實現(xiàn)就是不完整的。Howlongaretheseinteroperableversionslikelytobearound?Gravitatingtoward?多久這些可能在可互操作的版本嗎?會吸引protocolsstandardizedthroughopenprocessesmaymeanmoresupportforaprotocol,協(xié)議通過開放過程標(biāo)準(zhǔn)化協(xié)議可能意味著更多的支持，andhencealongerprotocollife.Oritmaynot.Don'tassumeOSPF,IS-IS,orEIGRPwill，因此協(xié)議時間較長?；蛘咚赡懿皇?。不要以為OSPF、是什么或EIGRP的意愿alwaysbearound—considerthesupporteachprotocolhasinreallife,howwidelyitsused,總是around-consider每個協(xié)議的支持已經(jīng)在現(xiàn)實生活中，有廣泛的使用，andhowlikelyitistobeusedinthefuture.和使用的可能性有多大。Howmuchareyouwillingtopayfortheabilitytousemultipleplatformsfrom?你愿意付多少能夠使用多個平臺multiplevendors?Standardsbodiescanonlypromisesolidstandards,not多個供應(yīng)商?標(biāo)準(zhǔn)機構(gòu)只能承諾固體標(biāo)準(zhǔn)，不是interoperability.Areyouwillingtospendthetimeandmoneyrequiredtoensurethebasic互操作性。你愿意花時間和金錢需要確保基本的functioningofeachprotocol,andthedesigneffortrequiredtoavoidfeaturesthatwon't每個協(xié)議的功能，避免功能所需的設(shè)計工作，不會的workacrossmultipleplatformsandimplementations?Savingatthebargainingtable跨多個平臺和實現(xiàn)工作嗎?保存在談判桌上whenbuyingequipmentisattractive,butremembertocountthecostsontheotherendof當(dāng)購買設(shè)備是有吸引力的,但請記住計算成本的另一端thebargainwhendecidingwhatprotocoltouse.當(dāng)決定使用什么協(xié)議討價還價。Intheend,youshouldchoosetheprotocolthatbestfitsyourneeds,takingfullaccountof最后,你應(yīng)該選擇最適合您的需要的協(xié)議，充分考慮thecostsandbenefits.Quiteoften,thedecisionwillswayinthedirectionofusingonly成本和收益。通常，這個決定將影響使用唯一的方向widelyimplementedstandards-basedprotocols,butarchitectsshouldnotgointothe廣泛實現(xiàn)基于標(biāo)準(zhǔn)的協(xié)議,但建筑師不應(yīng)該進入decisionprocessassumingthisistherightchoice.決策過程假設(shè)這是正確的選擇。HowEasyIstheRoutingProtocoltoConfigureandTroubleshoot?多么簡單的路由協(xié)議配置和解決？Whichprotocolismoredifficulttoconfigure?Whichismoredifficulttotroubleshoot?協(xié)議是更難配置?這是更難以解決？Clearlythesequestionscannotbeansweredinamoreobjectivewaythananyofthe顯然無法回答這些問題比任何更客觀的方法previousquestionswe'veasked.Butalthoughtherearenoclear-cutanswers,thereare之前我們問的問題。但是，盡管沒有明確的答案，有stillguidingquestionsyoucanaskinthisareatohelpyoudecidewhichprotocolfitsyour這個地區(qū)仍然指導(dǎo)的問題你可以問來幫助你決定哪些協(xié)議適合你networkthebest.網(wǎng)絡(luò)是最好的。Doyoutendtohaveroutingpolicyspreadthroughoutthenetwork,orconcentratedin?你往往有路由策略傳遍網(wǎng)絡(luò)，或集中在afewsmallplaces,suchasalongmoduleedges?BecauseEIGRPisadistancevector一些小的地方，比如在模塊邊緣嗎?因為EIGRP距離向量protocol,policycanbeimplementedalmostanyplaceinthenetwork,whereaslinkstate協(xié)議,政策可以實現(xiàn)幾乎任何地方的網(wǎng)絡(luò)，而鏈接狀態(tài)263263年protocols(IS-ISandOSPF)onlyallowpolicy,suchasroutefilters,tobeimplementedat協(xié)議(是什么和OSPF)只允許政策,如路由過濾器，實現(xiàn)well-definedpointsinthenetwork.網(wǎng)絡(luò)中定義的點。IsyournetworkbrokenintomanysmallerIGPdomainsconnectedthroughBGP?Do?是你的網(wǎng)絡(luò)分解成許多較小的顯卡域連接通過邊界網(wǎng)關(guān)協(xié)議嗎?做youuseasingleIGPend-to-endwithBGPasanoverlay,ornoBGPatall?Largerscale你使用一個顯卡的端到端邊界網(wǎng)關(guān)協(xié)議作為一個覆蓋，或沒有邊界網(wǎng)關(guān)協(xié)議嗎?更大的規(guī)模IGP-onlyorIGPend-to-endnetworkstendtobenefitfromEIGRPsabilitytoaggregateIGP-only或顯卡的端到端網(wǎng)絡(luò)往往受益于EIGRP聚合的能力andimplementroutingpolicyinmanyplacesthroughoutthenetwork.IftheIGPis和實現(xiàn)整個網(wǎng)絡(luò)路由策略在很多地方。如果顯卡isolatedintosmaller“pockets,”interconnectedwithBGP,orthenetworkscaleissmall分離成更小的“口袋”，相互聯(lián)系與邊界網(wǎng)關(guān)協(xié)議，或網(wǎng)絡(luò)規(guī)模很小enoughtoallowfortwoaggregationand/or“policypoints,”end-to-end,thenalinkstate足以讓兩個聚合和/或“政策點，”端到端，然后鏈接狀態(tài)protocolwillworkwell.協(xié)議將工作做好。Doyouneedtrafficengineeringthatrequiresanend-to-endviewofthetopology?你需要交通工程，需要拓撲的端到端視圖withintheIGPitself?Ifyou'replanningtousetheIGP'scapabilitytoperformconstrained在顯卡本身?如果你計劃使用顯卡的功能進行限制SPFfortrafficengineering(suchasMPLS-TE),thenalinkstateprotocolisyouronlyreal防曬系數(shù)為交通工程(如MPLS-TE),然后一個鏈路狀態(tài)協(xié)議是你的唯一option.Iftrafficengineeringwillbeperformedoffline,thenEIGRPisanoption.選擇。如果交通工程將離線進行,EIGRP是一種選擇。Whatisthenetworkoperationsstaffmorecomfortabletroubleshooting?Manypeople?網(wǎng)絡(luò)操作人員更舒適的故障診斷是什么?許多人findthedatabaseofalinkstateprotocoldifficulttoreadandunderstand.EIGRP's發(fā)現(xiàn)數(shù)據(jù)庫的鏈接狀態(tài)協(xié)議難以閱讀和理解。EIGRP的topologytabletendstobeeasiertoreadandgraspquickly,particularlyforoperations拓撲表往往是容易閱讀和掌握迅速，尤其是操作staffwhodon'tneedtoexamineorreviewroutingprotocolsdetailsonaregularbasis.員工不需要定期檢查或?qū)彶槁酚蓞f(xié)議細節(jié)。Whichprotocolproducestheleastcomplexfinalconfiguration?Thereisalmostnoway?該協(xié)議產(chǎn)生的最復(fù)雜的最終配置?幾乎沒有toknowtheanswertothisquestionotherthantobuildconfigurationsforthefeatures知道這個問題的答案除了構(gòu)建配置的功能you'llneedtodeployatseveralplacesthroughoutyournetwork,andmakeasubjective你需要部署在整個網(wǎng)絡(luò)的幾個地方，并作出主觀的callaboutwhichoneistheleastcomplex.調(diào)用哪一個是最復(fù)雜的。Combiningtheseconsiderationswiththeothersoutlinedinthischaptershouldprovidea將這些考慮與本章中概述的其他人應(yīng)該提供一個fairlygoodindicationofwhichroutingprotocolwillworkbestinyourenvironment.相當(dāng)不錯的路由協(xié)議將在您的環(huán)境中效果最好。WhichProtocolDegradesinaWayThatWorkswiththeBusiness?這協(xié)議降低了ThatWorks與業(yè)務(wù)嗎？Oneofthemostcommonmistakesinnetworkarchitectureistofailtothinkaboutfailure.網(wǎng)絡(luò)體系結(jié)構(gòu)中最常見的錯誤之一是未能考慮失敗。Therewill,inevitably,beatimewhenthenetworkfailstooperateasitshould,whether將不可避免地成為一個網(wǎng)絡(luò)無法運行時，是否duetotheweather,orequipment,orasimplemismatchbetweenthenetworkdesignand由于天氣,或設(shè)備，或一個簡單的網(wǎng)絡(luò)設(shè)計和之間的不匹配businessoperations.Whenthenetworkdoesdegrade,it'simportantforittodegradeina業(yè)務(wù)操作。當(dāng)網(wǎng)絡(luò)并降解，重要的是降低的waythatleavesasmuchofthecrucialpartsofthebusinessrunningaspossibleforaslong方式，讓盡可能多的關(guān)鍵部分的業(yè)務(wù)運行盡可能長aspossible.越好。264264年Forinstance,whendesigninganetworktosupportahigh-speedfinancialtrading例如,當(dāng)設(shè)計一個網(wǎng)絡(luò)支持高速金融交易operation,atotalfailuremightbepreferredtoapartialfailure.Itmightbebettertofailto操作，一個徹底的失敗可能會傾向于部分失敗。它可能是更好的失敗deliveranytrafficthantodeliversome(orall)trafficslowly.Inthecaseofaretailstore,交付任何交通比傳遞一些(或者全部)交通緩慢。在零售商店的情況下，however,itmightbebettertokeeptheretailcentersupandoperational,socustomerscan然而，它可能是更好的保持零售中心和操作，這樣客戶就可以continuebuyingthings,evenifretailanalyticscannotbecollectedinrealtime.繼續(xù)買東西，即使零售分析不能實時收集。Theprotocolthearchitectchoosesmustprovidefortherightfailuredomainseparationin協(xié)議架構(gòu)師選擇必須提供正確的故障域隔離therightplacesinthenetworktoseparatecriticalbusinessfunctionsfromtherestofthe網(wǎng)絡(luò)中正確的位置分離的關(guān)鍵業(yè)務(wù)功能networkwherepossible.Theprotocolthearchitectdecidestousemustalsodegradethe網(wǎng)絡(luò)在可能的情況下。架構(gòu)師決定使用的協(xié)議也必須降低leastundernetworkfailuresonthetopologiesandsectionsofthenetworkwherethemost至少在網(wǎng)絡(luò)故障在網(wǎng)絡(luò)的拓撲結(jié)構(gòu)和部分最多criticalbusinessfunctionsareperformed.執(zhí)行關(guān)鍵業(yè)務(wù)功能。WhichProtocolWorksBestontheTopologytheBusinessUsually協(xié)議效果最好的業(yè)務(wù)通常拓撲Builds?構(gòu)建？Withallthedifferentnetworkingtopologiesavailableinthecurrentworldofdesign,itcan與所有可用的不同網(wǎng)絡(luò)拓撲在當(dāng)前設(shè)計的世界,它可以seemthatmakingaprotocoldecisionbasedonthetopologyofyournetworkisboundto似乎做的協(xié)議決定綁定到基于網(wǎng)絡(luò)的拓撲結(jié)構(gòu)beaverycomplexproblem.Matchingtheprotocoltothetopology,however,isvery是一個非常復(fù)雜的問題。拓撲匹配的協(xié)議，但是，非常simple.簡單。First,youneedtodecidewhatthemostimportantpartofyournetworkisfroma首先，你需要決定什么是最重要的部分你的網(wǎng)絡(luò)的convergencespeedperspective.Nomatterhowlargeanetworkis,you'llalwaysfindthere收斂速度的視角。無論網(wǎng)絡(luò)有多大,你總是會發(fā)現(xiàn)isoneplaceinthenetworkwhereyoucannotworkaroundthecentralityofconvergence是網(wǎng)絡(luò)中的一個地方，你不能解決收斂的中心speed.Forinstance,inmanyfinancialorganizations,thetradingfloormightbecrucial,速度。例如，在許多金融機構(gòu)，交易大廳可能是至關(guān)重要的，whilethenetworkcore,orfinancialbranches,canoperatewithslightlyslower雖然網(wǎng)絡(luò)核心，或金融分支機構(gòu)，可以稍微慢一些convergencespeed.Foraretailorganization,individualstoresmightbecrucialfroma收斂速度。零售組織,個人商店可能是至關(guān)重要的convergencespeedperspective,whilethenetworkcoremighthaveenoughflexibilityto收斂速度的角度來看，雖然網(wǎng)絡(luò)核心可能有足夠的靈活性workaroundalessthanoptimalprotocolchoice.工作在一個低于最優(yōu)協(xié)議的選擇。Notetheremightbemorethanoneareaofyournetworkwhereconvergencespeedis注意可能會有不止一個的網(wǎng)絡(luò)區(qū)域收斂速度consideredcrucial.Inthiscase,youhavetomakeavaluejudgmentaboutbalancingthe被認為是至關(guān)重要的。在這種情況下，您必須對平衡做出有價值的判斷needsofeachpieceofthenetworkintermsofconvergencespeed.每一塊需要網(wǎng)絡(luò)的收斂速度。Afterthisdecisionismade,youneedtoexaminethetopologyofthatsectionofthe做出這個決定后，你需要檢查的拓撲結(jié)構(gòu)的部分networktodeterminewhatthecontrollingtopologyisgoingtobe—thepartofthe網(wǎng)絡(luò)來確定控制拓撲將是的一一的一部分topologythatisgoingtohavethelargesteffectoncontrolplaneconvergence.Thereare拓撲，控制飛機上融合產(chǎn)生最大的影響。有onlythreetypesoftopologiesyouneedtolookfor:afullmesh,ahubandspoke,orsome只有三種拓撲需要尋找:一個完整的網(wǎng)，中心輻射,或一些formofring.Thesethreetopologytypesarevirtuallyalwaysgoingtohavethemost形式的戒指。這三種類型的拓撲結(jié)構(gòu)都是幾乎總是最265265年impactoncontrolplaneconvergencebecauseofthewaysinwhichroutingprotocols影響控制飛機上收斂，因為路由協(xié)議的方式interactwiththem.與它們進行交互。Forinstance,iftheindividuallocationsinaretailenvironmentaretheareaofthenetwork例如,如果個人的位置在零售環(huán)境中網(wǎng)絡(luò)的區(qū)域whereconvergenceiscrucial,youwouldexpectalarge-scalehub-and-spokedeployment收斂是至關(guān)重要的，您所期望的大規(guī)模的輪輻式部署tobethemaintopologythatneedstobeaddressed.Forafinancialordatacentercentric主要需要解決的拓撲。金融或數(shù)據(jù)中心為中心的environment,afullmesh,orsomeformofcomplexpartialmesh,mightbetheprimary環(huán)境中，一個完整的網(wǎng)狀，或某種形式的復(fù)雜部分網(wǎng)，可能是主要的factorinchoosingaprotocol.Ifthenetworkcoreisthemostcriticalareafromacontrol因素在選擇協(xié)議。如果網(wǎng)絡(luò)控制是最關(guān)鍵的核心區(qū)域planeconvergenceperspective,thesetofringsthatmakeupwhatislikelyapartialmesh飛機融合的角度來看，套環(huán)組成可能部分網(wǎng)格mightbetheprimaryfactorinchoosingaprotocol.Whenconsideringthisstep,可能會在選擇一個協(xié)議的主要因素。當(dāng)考慮到這一步，rememberthatit'snotthelargertopologyyou'reseeking.Rather,it'stheprimary記住，這不是你正在尋求更大的拓撲。相反，它的主topologycomponentsinvolved—whetherrings,fullmesh,orhubandspoke.全網(wǎng)狀拓撲組件involved-whether戒指,或中心輻射。Afteryou'vedeterminedtheprimarytopologycomponentyouneedtoworkaround,you在你確定主拓撲組件需要解決，你的canexaminethewayinwhichdifferentprotocolsinteractwiththesetopology可以檢查不同的協(xié)議的方式與這些拓撲components,andthuschoosethebestprotocolfromanetworkconvergencestandpoint.組件，從而選擇最佳的協(xié)議從網(wǎng)絡(luò)融合的角度來看。Linkstateprotocolstendtoperformbetterinfullmeshnetworksthandistancevector?鏈路狀態(tài)協(xié)議往往表現(xiàn)得更好比距離向量在全網(wǎng)狀網(wǎng)絡(luò)(specificallyEIGRP),simplybecausetheyfloodthencalculate,ratherthancalculatinghop(特別是EIGRP),僅僅因為他們洪水計算,而不是計算跳byhopasinformationabouttopologychangesisbeingdistributed.跳的是分布式拓撲變化的信息。Distance-vectorprotocolstendtosupportlarge-scalehub-and-spokenetworksbetter?課程協(xié)議傾向于支持大規(guī)模中心輻射型網(wǎng)絡(luò)更好thanlinkstateprotocols,becausedistancevectorprotocolsonlycarryreachability鏈路狀態(tài)協(xié)議比，因為距離向量協(xié)議只攜帶可達性information,ratherthanreachabilityandtopologyinformation.Thisimmediateand信息，而不是可達性和拓撲信息。這直接和naturallevelofinformationhidinghelpsdistancevectorprotocolsperformbetterin自然的信息隱藏幫助距離向量協(xié)議中有更好的表現(xiàn)environmentswherereachabilityinformationisallthat'srequiredtomakeoptimal可達性信息的環(huán)境中都是必需的最優(yōu)routingdecisions.路由決策。Linkstateprotocolswillcreatemicroloopswhenconvergingonringtopologies,and?鏈路狀態(tài)協(xié)議時將創(chuàng)建微型接種環(huán)在環(huán)拓撲會合，和distancevectorprotocolswilldropsomeamountoftrafficwhenconverginginthesame距離向量協(xié)議將會下降一些交通融合時相同的environment.Choosingaprotocolfordeploymentinlarge