双活数据中心与灾备解决方案-技术部分

©2014VMwareInc.Allrightsreserved.双活数据中心与灾备解决方案-技术部分臧铁军@VMwareGCHCOE云架构师议程1同城双活技术方案2异地灾备技术方案2基于虚拟化技术的业务连续性解决方案概览资源池vSpherevSpherevSphere本地站点灾备站点•基于虚拟化层的异步复制•基于硬件设备的同异步复制•自动化应用切换管理•城域集群•应用感知的高可用性•关键应用零停机保护•在线迁移虚拟机，动态调配计算与存储资源•VMotionandStorageVMotion•高效的数据备份与恢复•可通过运行计划与脚本实现自动化操作资源池vSpherevSphereDev/TestDev/TestDev/Test灾难恢复本地高可用数据保护方案特点•与应用程序和操作系统无关•与硬件设备无关•完善的保护•简单，经济3议程1同城双活技术方案2异地灾备技术方案4双活数据中心在各个级别上全面保障可用性vMotion&DRSHA&FT服务器StoragevMotion，StorageDRS存储硬件热添加多网卡绑定存储多路径组件数据DataReplicationMetroCluster站点5双活数据中心总体架构双活存储集群站点A站点B延伸的vSphere集群200km•行为与单个vSphere相同•延伸距离最大200KM，通常小于50KM•通过VMwareHA与vMotion实现自动的DR保护•需要双活存储集群，如EMC的vPlex，NetApp的MetroCluster等6计算资源设计MakinganApplicationServiceHighlyAvailable•vSphereHA•vSphereAppHA8VMwarevFabric™tcServervSphereAppHAPolicy-basedProtectoff-the-shelfapps9FaultTolerancevs.HighAvailability•Faulttolerance–Abilitytorecoverfromcomponentloss–Example:Harddrivefailure•HighavailabilityUptimepercentageinoneyearDowntimeinoneyear993.65days99.98.76hours99.9952minutes99.999“fivenines”5minutesX10支持多vCPU的容错技术InstantaneousFailover4vCPU4vCPUvSpherePrimarySecondaryFastCheckpointing•FTsupportforupto4vCPUs•Protectionforhighperformance,multi-vCPUVMs•New,morescalabletechnology:fastcheck-pointingtokeepprimaryandsecondaryinsyncOverview•Protectmission-critical,high-performanceapplicationsregardlessofOS•Continuousavailability:zerodowntime,zerodatalossforinfrastructurefailures•FullyautomatedresponseBenefits11•Targetingcross-continentaldistances–upto100msRTTs•MaintainstandardvMotionguarantees•LeaderinVMflexibility功能简介•Permanentmigrationsbetweendatacenters•Disasteravoidance(DA)•DR/DAtesting•Multi-siteloadbalancing•FollowthesunBenefits长距离vMotionvSphere6.0支持跨三层网络和跨vCenterServer的vMotions12vCenterAvailability•RunvCenterServerapplicationinaVM•RunvCenterServerdatabaseinaVM•RunbothinsameVM?•ProtectwithvSphereHA–vCenterandDBVMrestartprioritysettoHigh–EnableguestOSandAppmonitoring–AppHAcanprotectSQLServerdatabase•BackupvCenterServerVManddatabase–Image-levelbackupforvCenterServerVM–App-levelbackupusingagentfordatabasebackup13网络资源设计双活数据中心网络架构物理二层(裸光纤)逻辑二层层叠网络/VPN二层分段VMVMVM二层分段VMVMVM二层网络扩展的二层网络(二层内容在数据中心互连链路上传递)站点A站点B二层网络15NSXvSphereMulti-SiteUseCasesNSXforvSpheresupports3differentMulti-SiteDeploymentModels1.VXLANwithStretchedClusters(vSphereMetroStorageCluster)2.VXLANwithSeparateClusters3.L2VPNAllsolutionsprovideL2extensionoveranL3network,enablingworkload&IPmobilitywithouttheneedtostretchVLANsLocalegressissupported,howeveritdoesaddcomplexityTheappropriatedeploymentmodelwilldependoncustomerrequirementsandtheirenvironmentNSX利用层叠网络实现双活数据中心双活存储vSphere城域存储集群数据存储1数据存储2vCenterServer三层网络站点A站点BVM1VM2VM3逻辑交换机A172.16.10.0/24VM4VM5逻辑交换机B172.16.20.0/24分布式逻辑路由器站点A边界网关上联网络A站点B边界网关上联网络B17VMwareNSXMulti-SiteSingleVC,StretchedClusterSolutionDetail•RequiresasupportedvSphereMetroStorageClusterconfiguration•InavMSCdeployment,storageisActive/Activeandspansbothsites.ExamplesofActive/Activestorageare:EMCVPLEX,NetAppMetroCluster(seeVMwareHCLformoreinformation)•StretchedclusterssupportLivevMotionofworkloads•UseL3forallVMkernelnetworks:Management,vMotion,IPStorage•AllmanagementcomponentssuchasvCenterServer,NSXManagerandControllersarelocatedinSiteA•LatencyandbandwidthrequirementsaredictatedbyvMSCstoragevendor,eg10msRTTforVPLEXwhichalsoalignswithvMotionusingEnterprisePlus•vMSCenablesdisasteravoidanceandbasicDisasterRecovery(withouttheorchestrationortestingcapabilitiesofSRM)•LossofeitherNSXComponentsortheDatacenterInterconnectwillresultsinafallbacktodataplanebasedlearningusingexistingnetworkstate.ThereforethereisnooutagetodataforwardingandwithoutvCenterServer,therearenoVMprovisioningormigrationoperations•NSXandvMSCarecomplimentarytechnologiesthatfitasweetspotforNSX(SinglevCenterServer)VMwareNSXMulti-SiteSingleVC,StretchedClusterVMwareESXiVMwareESXiVMwareESXiVMwareESXiSiteASiteBStretchedWorkloadClusterVMwareESXiVMwareESXiVMwareESXiVMwareESXiSiteASiteBStretchedEdgeClusterClusterConfiguration•vMSCenablesstretchedclustersacrosstwophysicalsites•InanNSXdeploymentManagement,EdgeandWorkloadclustersareallstretched•UndernormalconditionsallManagementComponentsruninaSiteAandareprotectedbyvSphereHA•TheyareautomaticallyrestartedatSiteBintheeventofasiteoutage.ThemanagementnetworkisnotstretchedandmustbeenabledonSiteBaspartoftherecoveryrunbook•Dependentondesign,NSXEdgeServicesGatewaysareeitheractiveinbothsitesorasinglesiteandcanalsoleverageHA•VMsintheWorkloadClustersareautomaticallyrecoveredVMwareESXiVMwareESXiVMwareESXiVMwareESXiSiteASiteBvCenterServerStretchedManagementCluster19VMwareNSXMulti-SiteSingleVC,StretchedClusterInavMSCenvironment,DRSisusedtobalanceresourceutilization,providesiteaffinity,improvedavailabilityandensureoptimaltrafficflowUseShouldrules,ratherthanMustasthisallowsvSphereHAtotakeprecedenceExampleDRSGroups,RulesandSettingsforNSXEdges:VMwareNSXMulti-SiteSingleVC,StretchedClusterNSXConfiguration(Option1-Preferred)•TransportZonespansbothSitesandVXLANLogicalSwitchesprovideL2connectivitytoVMs•DistributedLogicalRoutingisusedforallVMstoprovidecon