dubbo教程高級(jí)篇第20節(jié)消息中間件在分布式系統(tǒng)的作用介紹mq對(duì)比來源于

、MQ、主要關(guān)注前三個(gè)（標(biāo)紅高高中中中中低低中高高中中中低高低低多多中多中中少少中由于ErlangLinux高43方式CJMSPerl、.net等C+（不成python、java、php、.net等庫負(fù)載好有console實(shí)現(xiàn)無好無優(yōu)點(diǎn)戶端；缺點(diǎn)且對(duì)5.xActivemq不 erlang持amqpamqp的客優(yōu)點(diǎn)用（JMS的50臺(tái)左右，broker中；缺點(diǎn)mq中 等接的API下層MQ有多個(gè)mqnotify面mq這支持的ProtocolsThefollowingtableliststhesupportedprotocolsofeach0-8,0------------------STOMP-------InterfacesC------------Java--Java------------------------mq ark[文獻(xiàn)MQrabbitmqkafka傳統(tǒng)MQScenarioA:Enqueuing20,000messagesof1024byteseach,thendequeuingthemScenarioB:Enqueuinganddequeuingsimultaneously20,000messagesof1024byteseach.ScenarioC:Enqueuinganddequeuingsimultaneously200,000messagesof32bytesScenarioD:Enqueuinganddequeuingsimultaneously200messagesof32768bytesScenarioScenarioScenarioScenarioKafkavskafkarabbitmq的對(duì)比中，kafka工作是保證消息consumer是一直處于alive狀態(tài)去消費(fèi)消息的；broker重，需要記錄很多狀充分考慮消息堆積因consumer不alive狀態(tài)；考慮各個(gè)角色的分布式；為追求吞吐量設(shè)broker設(shè)計(jì)較輕，不保存consumer的消提供了豐富的的exchange、bindingqueuing等modelqueue、topic等routing都topicroutingtopicrouting對(duì)應(yīng)用透明的集群式但是集群對(duì)producer并不是完全透明，producer需要確認(rèn)消同這一點(diǎn)實(shí)現(xiàn)了消息的順序遞交特性（partition內(nèi)的消息是順序的mirrorqueue來masterslave通過replica來支持 master/slave的可以自動(dòng)切換（當(dāng)master支持，但需要人工切換masterslave消息量消息量消息量。mqmetaq，metaqlinkedinkafka（scala）java版本，并對(duì)其增加了事務(wù)的支持mqmetaq3.0kafka，其擅以替代傳統(tǒng)大部分MQ。。notify，meta2.x以前的所有隊(duì)列由于mq還比較新且沒有給出相應(yīng)的benark而mq社區(qū)主要活躍者均國(guó)內(nèi)人員，因此不便于直接對(duì)比mq和其他隊(duì)列。在這里用kafka和其他隊(duì)列進(jìn)行對(duì)比，mq的性能比kafka還要好一些。kafkavsrabbitmqinKafkaisageneralpurposemessagebroker,likeRabbItMQ,withsimilardistributeddeploymentgoals,butwithverydifferentassumptionsonmessagemodelsemantics.Iwouldbeskepticalofthe"AMQPismoremature"argumentandlookatthefactsofhoweithersolutionsolvesyourproblem.UseKafkaifyouhaveafirehoseofevents(100k+/sec)youneeddeliveredinpartitionedorder'atleastonce'withamixofonlineandbatchconsumers,youwanttobeabletore-readmessages,youcandealwithcurrentlimitationsaroundnode-levelHA(orcanusetrunkcode),and/oryoudon'tmindsupportingincubator-levelsoftwareyourselfviaforums/IRC.UseRabbitifyouhavemessages(20k+/sec)thatneedtoberoutedincomplexwaystoconsumers,youwantper-messagedeliveryguarantees,youdon'tcareaboutordereddelivery,youneedHAatthecluster-nodelevelnow,and/oryouneed24x7paidsupportinadditiontoNeitheroffersgreat"filter/query"capabilities-ifyouneedthat,considerusingStormontopofoneofthesesolutionstoaddcomputation,filtering,querying,onyourstreams.OrusesomethinglikeCassandraasyourqueryablecache.Kafkaisalsodefiniynot"mature"eventhoughitis"productionready".Details(caveat-myopinion,I'venotusedeitheringreatanger,andIhavemoreexposuretoFirstly,onRabbitMQvs.Kafka.Theyarebothexcellentsolutions,RabbitMQbeingmoremature,butbothhaveverydifferentdesignphilosophies.Fundamentally,I'dsayRabbitMQisbroker-centric,focusedarounddeliveryguaranteesbetweenproducersandconsumers,withtransientpreferredoverdurablemessages. WhereasKafkaisproducer-centric,basedaroundpartitioningafirehoseofeventdataintodurablemessagebrokerswithcursors,supportingbatchconsumersthatmaybeoffline,oronlineconsumersthatwantmessagesatlowlatency.RabbitMQusesthebrokeritselftomaintainstateofwhat'sconsumed(viaacknowledgements)-itusesErlang'sMnesiatomaintaindeliverystatearoundthebrokercluster.Kafkadoesn'thavemessageacknowledgements,itassumestheconsumertracksofwhat'sbeenconsumedsofar.BothKafkabrokers&consumersuseZookeepertoreliablymaintaintheirstateacrossacluster.RabbitMQpresumesthatconsumersaremostlyonline,andanymessages"inwait"(persistentornot)areheldopaquely(i.e.nocursor).RabbitMQpre-2.0(2010)wouldfalloverifyourconsumersweretooslow,butnowit'srobustforonlineandbatchconsumers-butclearlylargeamountsofpersistentmessagessittinginthebrokerwasnotthemaindesigncaseforAMQPingeneral.Kafkawasbasedfromthebeginningaroundbothonlineandbatchconsumers,andalsohasproducermessagebatching-it'sdesignedforholdinganddistributinglargevolumesofRabbitMQprovidesrichroutingcapabilitieswithAMQP0.9.1'sexchange,bindingandqueuingmodel.Kafkahasaverysimpleroutingapproach-inAMQPparlanceitusestopicexchangesBothsolutionsrunasdistributedclusters,butRabbitMQ'sphilosophyistomaketheclustertransparent,asifitwereavirtualbroker.Kafkamakesitexplicit,byforcingtheproducertoknowitispartitioningatopic'smessagesacrossseveralnodes,thishasthebenefitofpreservingordereddeliverywithinapartition,whichisricherthanwhatRabbitMQexposes,whichisalmostalwaysunordereddelivery(theAMQP0.9.1modelsays"oneproducerchannel,oneexchange,onequeue,oneconsumerchannel"isrequiredforin-orderdelivery).Putanotherway,Kafkapresumesthatproducersgenerateamassivestreamofeventsontheirowntimetable-there'snoroomforthrottlingproducersbecauseconsumersareslow,sincethedataistoomassive.ThewholejobofKafkaistoprovidethe"shockabsorber"betweenthefloodofeventsandthosewhowanttoconsumethemintheirownway--someonline,othersoffline-onlybatchconsumingonanhourlyorevendailybasis.Kafkacandeliver"atleastonce"semanticsperpartition(sincemaintainsdeliveryorder),justlikeRabbitMQ,butitdoesitinaverydifferentway.Performance-wise,ifyourequireordereddurablemessagedelivery,currentlyitlookslikethere'snocomparison:KafkacurrentlyblowsawayRabbitMQintermsofperformanceonsyntheticbenarks.Thispaperindicates500,000messagespublishedpersecondand22,000messagesconsumedpersecondona2-nodeclusterwith6-diskRAID/en...OfcoursethiswaswrittenbytheLinkedInguyswithoutnecessarilyexpertRabbitMQinput,soFinally,a KafkaisanearlyApacheincubator Itdoesn'tnecessarilyhavethehard-learnedaspectsinNow,awordonAMQP.Frankly,itseemsthestandardisamess.Officiallythereisa1.0proposedspecificationthatisgoingthroughtheOASISstandardsprocess.InpracticeitisaAsetofgenerallyavailable,widely-adopted,production-qualityAMQP1.0implementationsacrossthemajorreleases(QpidfromRedhat,RabbitMQ,etc.)won'texistuntil2013,ifever.Asanexternalobserverwithnoinsideknowledge,hereiswhatitlookslike:theworkinggroupspent5yearsonaspec,from2003to2008,culminatinginawidelyadoptedrelease(0.9.1).Thenasubsetofmorepowerfulworkinggroupmembersrewrotethespecbylate2011,compleyshiftingthefocusofthespecfromamessagingmodeltoatransportprotocol(sortoflikeTCP++),anddeclaredit1.0.So,wehavethestrangecasewherethe"mature"AMQPisthenon-standard0.9.1specificationandthe"immature"AMQPistheactual1.0standard.Thisisn'ttosuggest1.0isn'tgoodtechnology,itlikelyis,butthatit'samuchlower-levelspecthanAMQPintendedtobeformostofitspublishedlife,andisnotwidelysupportedyetbeyondprototypesando