今天我们会介绍一下为什麽我们的SRS基金需要一个讯件然后我们再会介绍一下一个比较普通的级面设置它是怎样构成的然后我们再会去完整介绍一下一个包含资予的实验性然后再会介绍一下怎样的设计,最后我们会展望一下未来的利用下一个公众我们先看一下,我们先看到一个比较大的现状第一个,我们先看一下栏豆子,我们的豆子里面为什么节点已经是超过一半了除了印尘之外,还有不同种类的材料,我们的包含资产业的版本还有一些这一份上万的级面设计,我们会展开扩展的,上万台节点的包含包括十万个节点的扩展 yummy我们再看看好多的种类,还有一个就是比較生命主機叫的,这变成是策略觉得不过互了固定 副指揮官的报道我们在实践中会有非常多的非常多的插槽很对应的 它的用户是自然自然是对我们上海的看到在这么多次元其实我们的KBS我们可以看到KBS对我们做成了或者是给我们的期望是它保持一个一致性 所有的可能性会将所有的一致性自然送到一个状态但是整个机子的一个所有自然我们必须问我们要如何监视它们的解释它们的解释同时就有很大的货带因为它们的一个和计它们的确很自然不如我们的一个我们的解释就是每一个期望只能管干嘛或者想想他们的建议會引起了一個問題甚至是一個事實的影響那肯定會有一個問題另外一個問題是因為我沒有想過這些資源已經沒有一個問題所以我明明想想沒有一個問題它的東西也還在機密那就是第一個是資源方式這樣會保證你的那些資源會去取得相關的資源然後還有是屬於那些既然資源還有證明它們屬於存儲資源這些把揮所謂導致其實你的有關資源的數據是第一階的還有這上面提到這些資源這來自周邊會有關資源的有關資源的有關資源的一些貨的資源有關資源的有關資源的有關資源的有關資源的有關資源的有關資源的有關資源的有關資源的有關資源的有關資源的有關資源的有關資源的還有有關資源的這上面提到有關資源的有關資源的有關資源的有關資源的有關資源的有關資源的有關資源的那這些資源會保證你的資源的ということ那你去哪裏去找之前的後 昨天 directions他們他們他們那我们需要一个什么样的一个询件所以我们 kind of 应该先按介绍一下但是 here I want to introduce three features of the expander那我们这个点是有上万台规模的同时它有十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模十万台规模我現在想要介紹一下第一種版本這是第一種版本的設計這是一個單機組件這是一個進程裡面的一些組件的試圖這就是我們經驗的過程我們經驗的過程是我們資源後的都會非常多大氣勢它的保持任何的態勢像外觀的氣變它是有的有的有的有的有的這就是我們要去做的這個就是它設計的然後我們要做一個有一個出發的機制這個就是第十個討論要的是有一個討論的機制然後我們要提供外觀的試驗去做出發或是它可以做出這裡外觀的這是我們所說的外觀的試驗這是我們在第一版本中使用的方式所以我們在外觀的試驗還有一個外觀的試驗我們也會使用外觀的試驗來展現資源資源的資源我們有興趣所以我們會進行第一個這個是起點它會它會去它的確診目的是這個任務對待任務對待是加一個任務可以是我們這個任務是搬回IT的那它把IT保護那這個IT就會去根據我們的Provider去取出去計算出結果當結果然後後面做展示這個這個我這邊說那麼,表現到變成一個很困難的問題雖然我們知道的問題並不能變得很重因為這粒任何考慮的問題能夠穩定得很長的時間所以,我們會很快地分析轉換考慮時,可以重溫我們會施加其他漲力然後,我們就得分析基本上,我們要創造一個新的版本我們其實想要基礎一些新的版本我們需要創造一個更新的版本現在我想要解釋一下Aitna can actually provide more training for us, it can be Canadian events,Aitna can be give hub and events.Cloud events is actually a general event model.And also, we do not want to only provide internal service,we want to provide a general framework to everybody.Of course, the potential model is not the same, but we want to provide a framework.So here, you can see based on the Canadian number,we actually establish new inspection models.Aitna, so the whole logic that actually does not change,providers do provide an update as the full inspection.Aitna talks here,Aitna already become a Canadian based point.Aitna, the event is actually a provisional trigger,Aitna, which is the most of the massive code.And now, I want to briefly introduce this process,this process of inspection.So suppose,Aitna probably has an event,so when the event happens,Aitna can actuallyenter the task to the task key.Aitna,so the task can subtract right into the connected event.Aitna,so when you receive the event,Aitna can actually compare it into the event.And the code is the result.And also,the cloud event of connecting will be sent back.And after sending back,Aitna is going to pass the message to Q again.And then,Aitna的acceptional extension,Aitna, which subscribes the result of the inspection.And then send it out into the acceptional system.So that means that if there is a problem,then the audience might during the inspectioncan ask them,or may be some special event that happensfrom the application.Aitna,we can easily,Aitna,we receive an industrial information.Aitna asks you,you can see.In the version 2,it basically solves a lot of open-top issuesof performance and scalability.So we can focus and make it as a frameworkto share with the company.So this is the very important introductionand inspection service.And the powerful of the inspectionis there is more to be done.And here,I only choose one inspection service.And this is just a beginning pointto our inspection system.Based on this beginning point,we can add a lot of feedbackfrom trust in our practices.We have an LSD for masodology.And my colleague,Chen Jie,will introduce you this pathology.Thank you.Thank you.And now I will hand overto introduce you the following part.And just now,Marty D had introduced youin the large-scale Kubernetes clustershow to practice the inspection.And there is a general inspection serviceto be adoptedto identify more of normal situations.And we also talked abouthow to introduce this inspection service.And today we had embracedthe seven speedsand we had also improvedthe performanceand the problem of scalability.So the inspection serviceis only at the beginning pointpoint of departure.And next,I'd like to introducehow to build a complete inspection systembased on that point of departure.So first,let me summarizewhat my colleague had observed.First,we would find thatwhat he is talking aboutis in a large-scale Kubernetes clusterswhat are the infrastructuresthat is the server,the network,the storage and the stabilitybetween all of those businesseshave a big gap.So where does this gap come from?We also know thatvery often we havea lot of monitoring systemsin the past years.And the monitoring systemcould identify some of the realitiesbut it cannot solveall of those problems.Usually,it can always solvea lot of the problemsand the problem of consistencycannot be solved by the monitor.So we are thinking aboutcan we use the inspectionto connect all of thoseinstructuresand to improve the stabilityof the businessesand to fill the gap.And we can providea comprehensive guaranteefor the stabilityin the Kubernetes systemsSo based on these motiveswe can think aboutthe featuresof the KubernetesThere are multiple problemsfor the monitoring systemsFor example,the Kuberneteswill collect the informationfrom the agentand there is a distortionof those informationand how to identifythe problems of themonitor itselfand we can useinspection serversto do that.So apart from the Kubernetesclusters and thethe monitoring systemswe have a lot ofdependent servicesand for examplemy colleagueHadutothe MDP and IPMand all of those systemswill have a data setand these datamore or less haverelationshipsof the Kubernetesdatafor example,the IPwhich is very commonfor all of thosesystems such asthe IPAMand OPSmost of thoseoperations andthe managementmaintenanceand whether it isinconsistencyinconsistencyor a conflictwe need to knowabout the root causeand we want tomaintain the consistencyso the consistencyusualhad been limitedto Kubernetesand if we wantto start froma larger scopeof thecloser loopwe canhave a betterdesignof the systemso based onthe Kubernetesplot-nativearchitecturetheorchestrationof the Kuberneteshas afoundamental changesHadutofor examplein terms ofthe APIwith a lot ofchangesandbasedon thesmallthe APIisimparativeandthe Kubernetesdestructureisdeclarativeand beforetheoperationalmaintenanceordesigncontainersbut nowit hadbeenevolvedto a flexiblecontainersin partso theoperationalmaintenanceor managementplothad beencompletely changedand beforewe onlyhad anadjustmentof theresource allegationand incommunitiesa lotof those jobshad beenpenetratedservicefindingand we alsohave a lot ofresourcesscansionsmanychangespart fromthe issuesI had pointed outwealsohas madesome manychangesandtakenagesmaybe in theinitial stagewe need tofocus ondifferent kind ofissuesand we need tothink aboutwhat are the keyissueswe need tofirstwithunderstandingof thisruleswe canhave a betterdesignand firstunderlayeris ourinfrastructureandwhat are theboundarieswhat infrastructureif you arerunning on acloudthen cloudis yourinfrastructurethestabilityof thecubinitiesrelight ontheinfrastructurethestabilityof thecubinitiesrelightandthen on thetopwe cansee thepart of thecontinuousandthis isother importantfor theinspectionsbut it willnotintologicsappreciationfor examplethere arelots ofthe e-commercetrade systemsandwe canhavedifferencesbetweenallergysystemandmonetarysystemif you areresponsiblefor themanagementthen you wouldbe veryfamiliar withthisconceptureso themonetarysystem willbase onthe matrixwhat arethe indicatorstomake decisionswhetherthissystem ismoresoverymodelogtype ofdatabutwhat abouttheinspectionsystemsforexampleyou shouldfirstthecenterforsomeexpectationsandthiskindofexpectationcanbereflectiveorindicatorsandsome oftheotherdependentdatassothis ismorefacing forthe targetandthe usersoinallergysystemno matterit's aboutmonetarysystemortheinspectionsystemwecanusewe have a lot of operatorsthosedevelopersresponsiblefor the mainnesses and the corefor the loading systemis to notifypeople aboutthe abnormalityso we hadtalked a lot aboutthe infrastructureand the boundariesof the infrastructuresand we also talked aboutthe differences betweenthe loading systemand the monitoring systemand then we cantake a look atthe components ofthe Kubernetes infrastructure systemthe first component is the nodethe dockerdpotchdpotchdisfinancialwordadapapacontainerperformanceso we need to runthe containerand this is the basisfor the successful scheduleand of the Kubernetesif there is a problematicenvironmentthen it's impossiblefor you to run the containersand when you havemodable versionsof the containersand you don't havethe right environmentfor thatit's impossiblefor you to run themappropriateso this is themost coreinspectionand the secondis theKubernetescorecircuitand we canhave theautomaticsof the coresso the corecircuitis aboutthe singlemachineand the EDCDso becauseof the corecontactwe cangenerate someof the containersand we needto think aboutwhetherand I hadbeenemphasizingon the externaldependent systemsand we also needto havebrown inspectionfor thatand in theroundtime of the containerwe also needto havethe inspection sessionsthe networkingand thestoragewhether theyhad beenmeetingyour expectationsand thethird partis aboutthe incidencesand Kuberneteshave alot of incidencesand theseriesfor thearrays好那那当然我说我们说刚才说的就说我们已经签个这个选减的办So now we knowthe scopeof the inspectionsand we knowthe valuesof the inspectionsand actuallywhen thereissome abnormalitydetectedabout inspectionsit is avery valuablethatso just nowI hadtalked aboutthe importanceof theautomalities identifiedabout inspectionsand they canhelp us tomake the decisionsand tohave more analysisand youcould knowthe underlyingproblemsand thenyou canexcludethose problemsand tofixthose issuesand moreimportantwhen thereis an issuewe canyou cansay thatyou are wellwe canuse the controllerto fix itbut in alargest scaleclusterit's impossiblefor youto do itcompletelymanuallyso we wantto do itautomaticallyand theninspectionand fixingshould beclackedand integratedso weneed toform aclosed loopso howjust nowhavingemphasizingon the provider of the dataand howto provide serviceswith the providerand thecluction of the datawhich you can seeat the bottomwe are usingthe Kubernetescommunityopen sourceexportmagnetismand the Kubernetesexportcancreatenot botand eventand we havethe metricsand we havethe standard longmachinesupplicersand theWistos datawe need tothink about storagebecause the formatof the dataare differentand we havea huge amountof the dataand themetrics will beexisted in aTSDBand the eventwill bestoredas welland themetadatasuch asnotand partwill bestored in adatabaseso thatwe canbetterwe canhavethe inspectiondiagnosisand thecalculationwhen there isan abnormalitywe canrun themonitoringand tofix itso wehave tofix enginesand tohave therecoverysowith somany capabilitieswe canprovidethe APIto connectto thetopproductsandtheself-healingwe needthesystematicalmanagementcapabilitiesfor exampleI cangive youself-healingand we canuse multiple toolsto fix itbut actuallyself-healingis a very dangerousthingfor examplewhen thecontainer has aproblemthenyou need todeletethe previouscontainerby a bugwhat should you dothe whole clusteris in dangermost dangeris the thingis that you had adeleted the whole clusterand this is quite riskyso you have tohave thesecurity measureswe havethe commonadoptionscommon practicesetc.the throttlingand thenyou alsoneed to have thefuse launchso you need topreventthe furtherunnecessary deletionsand you shouldalso have thebackupsnowin our companywe havevery comprehensiveself-healing mechanismswith thatyou can bettercontrolthe executionof self-healingand controlled riskso justnowwe had talkedabouthow tocreatea closed loopinspection systemandfinallythe purposeof theinspectionis toenablethe usersthe betterwrong systemare our usersso how to empowerthe usershow toallow them toconstantlyinctoratethe developmentsystemsand I cangive you an exampleand we usepromise useas a basisfor monitoringandif youhave to beusingpromise usebeforethen youknow thatwe canuse theinspection systemand it's the samefor the inspection systemfor examplewe can developour ownproductsor a setof the productsand thenwe can haveconfigurationand managementfor the clusterswe havemultipleusersmany usesand we havemultiplecategoriesof theclustersfor theusers and someof themaintenancethe systemand by themselveswe needthis itembut it's not neededin anotherclustersothe nominatedvalueor the thresholdfor each clusterare differentwe need a moreuser-friendly wayand alsoask my colleaguejust a mentiontomake an inspectionframeand workwithoutwithoutwithoutwithoutwithoutwithoutwithoutwithoutwithoutwithoutwithoutwithoutwithoutwithoutwithoutwithoutwithoutwithoutwithoutwithoutwithoutwithoutwithoutwithoutwithoutwithoutwithoutwithoutwithoutwithoutwithoutwithoutwithoutwithoutwithoutwithout我们的质疑网绸扰数量确实是让我们来运用国内的质疑网绸扣数量。我们的质疑拔除质疑也不在通用网公司,而我们的质疑也不在通用网公司,与我们的质疑网线通过质疑。而我们把我们的质疑扣数量及质疑的能量,我们有很多的观察系统因为我们经常在围检查的和发挥支援的Q&A因为我们的客户问我们为什么我的围检查没有工作所以他们可以在我们的围检查解决我们的围检查的结果所以他们可以进行进行解决的状态还有我们的所有围检查的和我们的围检查的我们可以在围检查的和发挥支援的就是这个事情就是在围检查的这是我们 在围检查的我们认为围检查的有看看这一种水平而在围检查的我们可以在道路上们可以在道路上呈现或是在围检查的和发挥支援的我们可以在道路上呈现或者是在围检查的和其他的数量在并在视频上如果有车厂或有一个在视频上的中辆可以从视频上出现我们会给予很多服务然后可以在每一个系统中和各种各样的不同的方式我想谈过的将来其中一个项目不是所以這是一種非常重要的問題還有更多關於預期的未來我們很希望在此我們可以使用預期訊號進行訊號進行訊號進行訊號然後在訊號內進行訊號資訊這個會是訊號進行訊號所以在此我們是想要進行訊號進行訊號進行訊號的訊號進行訊號我們也需要去做一個獨補的主席也也要對應我們的管理協調我們可以保持更多的控制度所以下個時間我們要進行參考這個案件A.I.O.P.O.S 指控自行充弄的預算系統所以這就是我的分享謝謝大家 謝謝大家