资源描述
,Click to edit Master title style,Click to edit Master text styles,Second level,Third level,Fourth level,Fifth level,*,Click to edit Master title style,Click to edit Master text styles,Second level,Third level,Fourth level,Fifth level,CONFIDENTIAL,*,单击此处编辑母版标题样式,单击此处编辑母版文本样式,第二级,第三级,第四级,第五级,*,*,单击此处编辑母版标题样式,单击此处编辑母版文本样式,第二级,第三级,第四级,第五级,*,*,VMware软件定义存储解决方案v32,VMware软件定义存储解决方案v32VMware软件定义存储解决方案v32,目录,存储发展趋势,什么是软件定义存储,Virtual SAN,竞争分析,Virtual SAN vs.Nutanix,Virtual SAN,实测报告,Virtual SAN,资源,&,反馈,2,注意事项,战略上藐视友商,战术上重视,尽量不主动地针尖对麦芒地比较,要用组合拳。但要了解友商的技术细节,知己知彼;,不主动谈友商的劣势,因为你很难比友商更先了解它的最新信息,如果没有深入了解其技术细节,或者不知道其最新信息,你可能会失去用户,/,合作伙伴的信任,谈存储发展趋势,(,分布式、软件定义代表未来,),用,Solution(,组合拳去,),竞争,(,将,VSAN,置于,Hypervisor,SDDC,云和自动化之下讨论,),VSAN,是,vSphere,的一个高级功能,vSphere +VSAN;vRealize+VSAN;Horizon View+VSAN;NSX+VSAN?,数据保护,+VSAN;,如,vReplication,VDP,VDPA,常规应用,+VSAN;,如,Oracle,SQL Server,Exchange,SharePoint,等,),行业应用,+VSAN;,金融、电信、教育、医疗、政府、制造业、云提供商、,战术上,将友商分为两大类:传统存储,(,外置磁盘阵列,),、其他,SDS,3,存储发展趋势,4,2011,2020,年,,IDC,的“数字宇宙”报告,1,、数据量将增长,50,倍,而,IT,人员仅仅增长,1.5,倍;,2,、非结构化数据约占未来十年数据产生量的,90%,;,有的时候,用户的计算量并没有增加,但是数据量增加了,需要更多的存储空间;,5,全球,IT,预算持续,节省成本真的很重要!,*Gartner 2Q2015,全球,IT,支出预测,;2015,年甚至为负增长,:-5.5%,(实际,2.5%,),;,SanDisk发布世界上第一款,4TB SSD,;,2015年3月,SanDisk推出的InfiniFlash,单卡容量达,8TB,的闪存卡(64个),;,SanDisk,计划,2016,年发布,16TB SSD,;,7,参考,:SSD,磁盘高性能低功耗,349 x 15K HDD,90,000 IOPs,11 x SAS SSD,88,000 IOPs,1 x PCIe SSD,90,000 IOPs,Price/Performance:$2.14/IOPS,Power:2443w,Price/Performance about$0.72/IOPS,Power:66w,Price/Performance about$0.2/IOPS,Power:13w,Random IOPs:70%Reads/30%Write;8K transfer;Q-Depth=4:2.5”,15K SAS,IOPs=258/Drive,Random IOPs:70%Read/30%Write;8K Transfer;Q-Depth=4:SAS SLC SSD,IOPs=8000/Drive,Random IOPs:70%Read/30%Write,8K Transfer:,IOPs=90,000/Drive,每块,15K HDD,,,250,多个,IOPS,每块,SAS SSD,,,8000,个,IOPS,每块,PCIe SSD,,,9,万,IOPS,存储发展趋势:,SDS(,软件定义存储,),10,存储发展趋势:,SDS(,软件定义存储,),Hyperconverged Storage Market for Google,Facebook,Yahoo etc.,Hyperconverged Storage Market for Commercial Enterprise.,Gartner,史上罕有的,为某单一产品做出分析和预测,“截至,2017,年,将有三分之一的中端市场组织至少会为其总存储容量的,30%,部署,VMware Virtual SAN.”,IT),不断增长,低效,业务停顿,难以满足,SLAs,故障诊断和解决的时间漫长,安全与合规的问题,复杂的,竖井式的,IT,架构,高额的,C,apex(,初始资本支出,),和,O,pex(,运营支出,),操作低效,VMware,对软件定义数据中心的描述,Software Defined Data Center(SDDC),抽象,自动化,池化,计算,存储,/,高可用,网络,/,安全,管理,Traditional IT pain points,New IT pain points,Vmware,的,SDDC(,软件定义的数据中心,),实现了变革,*Claims being validated by the Taneja Group(final numbers expected August,2013).,Inflexibility,Fear of Lock-In,Inefficiency,Downtime,超过,500,家,ISV,解决方案的支持,支持,80,种,OS,多种选择,Any AppAnywhere,高效,Capex(,初始资本支出,),减少了,75%,,,Opex(,运营支出,),减少了,56%*,Cloud ServiceProvider Economics,可控,Tier 1,应用的停机时间减少了,36%*,Cloud onYour Terms,敏捷,IT,生产率提高了,67%*,Apps atBusiness Speed,软件定义的数据中心解决了,IT,痛点,VMware,对,SDDC,和,SDS,的描述,VMware,做为,SDS,概念的创造者,对,SDS,定义如下:,软件定义的存储产品是一个将硬件抽象化的解决方案,它使你可以轻松地将所有资源池化并通过一个友好的用户界面(,UI,)或,API,来提供给消费者。一个软件定义的存储的解决方案使得你可以在不增加任何工作量的情况下进行纵向扩展(,Scale-Up,)或横向扩展(,Scale-Out,)。,软件定义存储是,VMware,软件定义数据中心的五大组成部分之一。,VMware,认为,软件定义的数据中心,是,IT,演变的下一个阶段,是迄今为止最有效、恢复能力最强和最经济高效的云计算基础架构方法。,SDDC,方法论将对存储、网络连接、安全和可用性应用池化、抽象化和自动化,,整个数据中心由软件自动控制,。基础架构提供的服务将聚合起来,并,与基于策略的智能调配、自动化和监控功能结合在一起使用,。应用编程接口和其他连接器支持,无缝延展到私有云、混合云和公有云平台,。,17,什么是软件定义存储,(Software Defined Storage),全球网络存储工业协会,(SNIA),认为,,,SDS,需要满足的是:,提供自助的服务接口,,用于分配和管理虚拟存储空间;,应该包括如下功能,:,自动化,Automation,Simplified management that reduces the cost of maintaining the storage infrastructure.,标准接口,Standard Interfaces,APIs,for the management,provisioning and maintenance of storage devices and services.,虚拟数据路径,Virtualized,Data Path,Block,File and Object interfaces that support applications written to these interfaces.,扩展性,Scalability Seamless ability to scale the storage infrastructure,without disruption to availability or performance,.,传统方式,手动传送数据需求,SNIA,观点,注:,什么是virtual storage devices?,Block Storage Logical Units,File system Shares,Object Containers,“storage administrator”defined and implemented.,让应用通过元数据来请求相应的,数据服务,,包括如下:,空间部署,数据保护,(,快照、克隆,),,数据高可用,(,容灾、双活,),,性能,安全等,SNIA,观点,理想方式,通过元数据传送数据需求,SNIA,观点,(,理想方式,),SDS,总揽图,举例,:,Amazon,EBS,云存储服务价格列表,Virtual SAN,竞争,分析,23,竞争分析讨论的大前提,vSphere用户,或者计划购买vSphere的用户,我们承认,VSAN,只在单一,Hypervisor,上运行,是我们的劣势,,但是同时也是优势,将更多话题引向优势面来讨论。,24,Magic Quadrant for x86 Server Virtualization Infrastructure,2014,At least 70%of x86 server workloads are virtualized,the market is mature,A poll taken at Gartners Data Center Conference in December 2013 showed that more than,90%,of attendees considered vSphere their primary hypervisor,70%的 x86 服务器工作负载实施了虚拟化,25,越来越多的关键业务应用运行在虚拟化环境里,VMware Customer Survey,26,VMware,拥有非常广泛的存储合作伙伴体系做为,Virtual SAN,和,Virtual Volumes,的强大后盾,vSphere,Virtual,SAN,基于存储策略的管理,支持,VVOL,的阵列,VMware,软件定义的存储,vSphere,Virtual,Volumes,迄今为止,最接近,SNIA,理想方式的,SDS,方案,-VMware SDS,SAN/NAS,SAN/NAS Pool,Virtual Data Plane,x86 Servers,Hypervisor-converged,Storage pool,Object Storage Pool,Cloud Object Storage,Virtual Data Services,Provisioning,Performance,Protection,Storage Policy Based Management(SPBM,基于存储策略的管理,),Virtual Volumes,Virtual SAN,vSphere,Policy-driven Control Plane,28,不要小瞧了,SPBM(,基于存储策略的管理,),SPBM,助力实现:存储即服务,(*Storage as a service),VAAI,,,VASA,出现后,开始逐步改变了这一现状。,现在,我们知道借助,VASA,,,vSphere/vCenter,已经可以知道存储的许多特性;如果存储支持,VMware vVol(Virtual Volume),,这一技术还能帮助存储管理员在存储,GUI,界面里,了解到虚机的详细情况,例如有哪些虚机连接到这台存储上,这些虚机分配多大空间;每个虚机有哪些,VMDK,,大小和使用空间多少,有无快照,内存、,SWAP,占多少。这样存储也能感知前端虚机,/,应用了,这在以前几乎难以想象。,3,、颗粒度细,以虚机为中心(也即,App-Centric,)进行存储资源的调度和管理。,以虚机,甚至每个虚机下的,VMDK,,这么细小的颗粒度,在存储中部署并被管理起来;利用率高,部署、管理和变更(如在线调整存储策略)都很方便。,30,创建虚拟机的存储策略,31,创建新虚机时,可选存储策略,32,变更存储策略,传统存储手动操作,vs.VMware VSAN,的自动化,传统存储需如下步骤,从原有数据存储迁移虚拟机,删除所涉及的,LUN,或卷,创建一个满足新的存储要求的新的,LUN,(不同的,RAID,等级,),重新以,VMFS,格式来格式化,LUN,将虚拟机重新迁移回新的数据存储,VSAN,仅一个步骤,(,可有两个选择,),修改现有的关联到该虚拟机的存储策略,创建一个全新的策略,再将新策略关联到该虚机,33,Automating Storage Selection with vCAC(vRealize)and VSAN,34,Automating Storage Selection with vCAC(vRealize)and VSAN,35,Virtual SAN,可简化并自动执行存储管理,从单个自调节的数据存储满足每个虚拟机的存储服务级别要求,基于存储策略的管理,Virtual SAN,共享数据存储,vSphere,+,Virtual,SAN,SLA,软件自动,控制服务级别,不再需要,LUN/,卷!,基于应用需求的,策略集,容量,性能,可用性,按虚拟机,设置的存储策略,36,如何应对外置磁盘阵列友商的竞争,?,一、更快,弹性扩展,可预测性能;闪存加速,性能稳定、延时低(小于,1ms,);,二、更省,降低,TCO,;可按需支付;部署、管理简单,自动化程度高,无需存储专家管理;,三、更简单,两次单击,即可部署完成;与,vSphere,紧密集成,界面统一,管理简单。基于策略的驱动,能自动调优;能在线线性扩展;,Tip:,Partner SE(,大,OEM,厂商,),谈论,SDS,时,尽量不用或者尽可能少用“,不再需要 LUN/卷!,”字眼,不谈论与外置磁盘阵列友商的竞争,避免引起对方反感。,37,以设备为中心的孤立小环境,静态服务等级,僵化的调配,缺乏精确控制,频繁的数据迁移,耗时的流程,自动化程度低,对请求响应慢,复杂的流程,非商用,利用率低,超额配置,昂贵的专用硬件,APP,或,VM,APP,或,VM,APP,或,VM,客户目前在传统外置存储方面面临着一些挑战,38,传统外置存储,vs.,软件定义的存储,全新控制层,从以硬件为中心转向以应用,/,虚机为中心,全新数据层,从专用硬件转向行业标准硬件,软件定义的存储,(SDS),传统外置存储,策略驱动的自动化,跨阵列的通用性,动态控制,Server,SAN,经过闪存加速,分布式,存储控制器普遍,X86,化,所以,SDS,在,RAS,上与传统存储相差不大;,39,价格比较:,Virtual SAN,使得,TCO,最多降低,50%,资金开销,服务器端经济优势,减少前期投资,精确的扩展,利用存储硬件成本趋势,提高资源利用率,线性可预测成本曲线,运营开销,提高管理员的工作效率,简化预算,无需专业技能,低至,0.50,美元,/GB,2,低至,0.25,美元,/IOPS,OPEX,降至,原来的,1/5-1/2,4,TCO,最多,降低,50%,5,每个桌面低至,50,美元,1,1.,完整克隆,2.,可用容量,3.,根据,2013,年市场定价估算的资金开销(包括存储硬件,+,软件许可证成本),4.,资料来源:,Taneja Group,5.,与传统的共享存储解决方案相比较,40,价格比较,前提是,vSphere,虚拟化环境,+,全闪存阵列,查看业界全闪存阵列价格比较,查看,Excel,表格:,Price Comparison-pye ver2.0,VSAN 13K vs.others 6080K,41,知识中心有,VSAN6.0,招标参数,v1.1,例如指标要求,提供基于存储策略的管理机制,根据业务应用的需求,可以通过变更存储策略进行在线的动态调整业务应用正在使用的存储资源,需要注意的是,那些支持,VMware Virtual Volumes(,简写为,vVol),的存储产品,应该认为他们支持基于存储策略的管理,不建议向用户,/,合作伙伴深入剖析,Virtual Volumes,的技术原理,42,如何应对友商的跨异构存储的存储虚拟化方案,?,跨磁盘阵列的存储虚拟化(也即异构存储的管理,互操作性),比较知名的有:,EMC VPLEX,、,IBM SVC,、,HDS VSP,,以及,Symantec Storage Foundation,。他们都能或多或少的将其他厂商的存储纳入自己的存储平台之下进行管理。但这种管理,也,只是将异构存储的逻辑卷做为一个外来设备使用,,把它视为一个普通的容器,不知道它能提供多大的性能。同时也丢失了异构存储内嵌的丰富的软件特性,例如快照、容灾等,;,存储虚拟化的缺陷,:,编程接口不会轻易对其他友商开放;服务扯皮,推诿责任;管理复杂度并未降低多少;,服务扯皮的举例:,“曾经有用户想用,HDS,的,USP,管理,EMC,的,CX,系列磁盘阵列,结果,EMC,工程师跟用户讲磁盘阵列的兼容列表上没有,HDS USP,,拒绝提供服务。还有一次用户实测用,NetApp,的,V3000,管理,IBM DS,系统磁盘阵列,发现性能低的离谱。结果,NetApp,和,IBM,的工程师都说不是自己的问题,让对方改设置来兼容自己”,43,vSphere,也是一种存储虚拟化的方案,其实,Hypervisor,本身也在用户不知不觉的过程中完成了存储虚拟化或池化的工作,这是更高级的池化,能够在虚机,/,应用级别去全局的、统一的、灵活的分配和协调存储资源。以,VMware,为例,,vSphere(,池化那些尚未支持,Virtual Volume,的外置阵列,),将存储资源池化后,即可借由,DRS,、,Replication,或,SRM,、或是免费的,VDP(Data Protection),实现虚机存储资源的动态调配,或是虚拟磁盘的备份、归档和容灾,,同样实现了异构阵列的利旧,。,44,VSAN,如何应对其他,SDS,友商的竞争,?,Virtual SAN,已嵌入,vSphere,内核,CPU,占用少于,10%,便于管理,无需安装和管理单独的虚拟设备,无单点故障,提供最短的,I/O,路径,与,vSphere,和,VMware,产品体系,无缝集成,vSphere,Virtual,SAN,已嵌入,虚拟机内核,虚拟,存储设备,.,+,45,市场上都有哪些,SDS,厂商?,最有可能遇到的友商是位于,Data Plane,里的,超融合架构,(HCI),如,Nutanix,、,EMC ScaleIO,、,Maxta,、,SimpliVity,、,Scale Computing,、,Pivot3,、,;,国内有:华为,FusionStorage,、,志凌海纳,SmartX,;,开源的有,Open vStorage,(类似,Nutanix,架构);,46,Virtual SAN,与,VMware,产品体系深度集成,VSAN,集成,High Availability,vMotion,Data Protection,vSphere Replication,等功能,vMotion,vSphere,HA,DRS,Storage,vMotion,vSphere,VMware,View,虚拟桌面,vCenter,Operations,Manager,vCloud,Automation,Center,IaaS,云环境运维和自动化,Site,Recovery,Manager,灾难恢复,站点,A,站点,B,基于存储策略的管理,快照,链接克隆,vSphere,Data,Protection,vSphere,Replication,数据保护,47,同城双活数据中心:,VSAN For Metro Cluster,VSAN,是最经济简便的双活数据中心存储解决方案,只需要在两个双活数据中心部署,VSAN,并通过高带宽低延时网络相连,与,vSphere HA/FT,,,vMotion,等技术紧密集成,RPO=0,,,RTO,分钟级,站点,A,生产站点,A,生产站点,B,仲裁站点,Virtual SAN for Metro Cluster,站点,C,站点,B,vmdk,witness,vmdk,witness,高带宽,1,0-40G,低延时,1,-5ms,低带宽,0.2-1M,高延时,2,00-,1,000ms,vmdk,vmdk,48,更多与集成的相关介绍,可参考藏铁军的,VSAN Exec,49,与其他,SDS,的比较,重点分析,Nutanix,的即可,50,Virtual SAN vs.Nutanix,注:,VSAN,最突出的优势是,,1,、内嵌在,vSphere,里,,I/O,路径短,延时低;耗,CPU10%,;,2,、与,vSphere,高级功能,如,HA,vMotion,DRS,等紧密集成;,3,、管理简单,仅单击两下鼠标即可部署;,4,、而且具备自动化功能,能够被,vSphere,或云平台,(,如,vRA),调用,实现软件定义数据中心所需的快速简单、敏捷交付,自动部署存储资源;,下面介绍一些技术细节上的差异;,51,Tips,不要拿,EVO:RAIL,与,Nutanix,相比较,因为,EVO:RAIL,配置相对低端些,VSA,方式导致的潜在问题(最好的方式是嵌入,Hypervisor,内核),来自一位,Nutanix,用户的使用反馈:,安装部署很方便,但是迁移维护出故障,将某个,Host,进入到维护模式之前,需要关闭,CVM,关闭,CVM,,出现无法访问,NFS datastor,,也不能进入,Nutanix Prism,管理界面,用户最后做出的结论是:,It is my opinion that although this hyper-converged platform is extremely simple to initially configure and deploy,the resiliency of the overall solution has,a long way to progress in order to fulfill the requirements of Enterprise adoption.I do not recommend Alexion moving forward with utilizing Nutanix-based virtualization in any capacity.,52,Nutanix,技术细节,耗资源多:,CVM至少要24GB内存,8个vCPU。若有去重压缩,需32GB内存;所有CVM都从SSD启动。,重删和压缩有局限:,1,、,采用MapReduce Dedupe做去重和压缩,以下两者不同时并存:(1)压缩有Inline和Post Process;,(2)去重有Performance(只减少内存消耗,不减少容量)和Capacity;,2,、,只在12GB(VMDK)里去重,适用于VDI(Image);,3,、,VDI才有必要重删,(备注:全球,60%,用于,VDI,),。VAAI clone和Linked clone环境不建议重删。,提防,Nutanix,上千个节点集群的提法:,Nutanix,实际部署最多,50,多个节点,但多数容量不够;国内有,16,个节点的集群;,NDFS,支持多个,Container;,VSAN,可以借助空间预留、缓存预留达到类似效果,又不失共享的好处;,Nutanix,的两个,10,Gb,是,Active/Passby,的;,53,Nutanix,的存储池也只是给,Hypervisor,集群的虚机使用的,mount the container as a NFS datastore,ESX,hosts.,mount the container as a SMB volume on,Hyper-V,hosts.,KVM,on Nutanix uses iSCSI and NFS for storing VM files:,1)iSCSI for VMs,2)NFS Datastores for Images,NFS datastores are used to manage images which may be used by multiple VMs,such as ISO files.,54,Nutanix Controller VMs Can Easily Use 25-50%of CPU,Nutanix,的虚拟存储控制器,(CVM),运行在虚机上,每个,Host,一个,CVM,,该,CVM,控制其他所有应用虚机的,I/O,访问,很累的。,下图显示,仅仅运行,IO,测试的应用虚机,占用,CPU,和内存都很少。然而四个,Host,的,Nutanix CVM,都非常繁忙,,CPU,消耗都在,14000MHz,以上,内存消耗在,73%,以上。,Nutanix CVMs are“busy”handling IO,User VMs are small and not using much CPU or RAM;running IO tests,55,Controller VMs Very Demanding,After migration,workload continues to drive load on previous CVM,I/O transitions to local CVM over time,After vMotion CVM spikes,Slowly Stabilizes,Note the Mem usage all reserved,Nutanix SEs often add vRAM beyond the 16GB shown here to CVM,CONFIDENTIAL,56,Nutanix,的数据本地性并没有什么优势,因为万兆网络延迟远低于,SSD,延迟,,而且,Nutanix,声称,:,Nutanix,确保,data locality(,数据本地性,,i.e.,存储与计算在同一个节点上,),,因此,Nutanix,比,VSAN,更快,.,真的吗?,Sounds nice but IT DOESNT MATTER!(at least not for VSAN),Writes,(写,IO,),must,occur on two different nodes so not local anyway(Same as VSAN),Reading from local disks,should be,more efficientbut on 10GbE the network isnt a huge bottleneck,10GbE TCP/IP latency:,100 microseconds(,微秒,),round-trip,PCIe flash latency:,1 millisecond(,毫秒,),(注意:万兆网络之间的来回延迟仅有,PCIe SSD,的十分之一!),Disk storage:,5 milliseconds(,毫秒,),10 GbE networking add some latency but this is NOT the bottleneck in the Nutanix system,its the CVM!,瓶颈不在于万兆网络,瓶颈在于,Nutanix,的,CVM,Key questions to consider about Data Locality:(,备注:,Nutanix,的数据本地性只是帮助读,IO),Write IO is slower than Read IO,and by design must be remote,so how does data locality help here?,(,写,IO,慢于读,IO,,而且必须跨节点做同步,因此,,Nutanix,的数据本地性对性能并没有多大的帮助,With Nutanix read IO should come from the local node what happens if the CVM is busy or a vMotion occurs?,如果,CVM,非常繁忙,或者,vMotion,发生了,,Nutanix,的本地读就没有优势了,10 GbE networks add only slight overheadso does it matter where the IO actually comes from?,万兆网络只是增加了非常小的延迟,因此,读,IO,来自于哪个节点,无关紧要,CONFIDENTIAL,Nutanix Data Locality:How It Works,Nutanix IO flows:,Reads:(,shown in green,),VM read request goes through CVM,Then thru vSphere,To datastoreshopefully SSD,Then back to VM,vSphere,Read request flow,Writes:(,shown in blue,),VM write request goes through CVM,CVM sends thru vSphere,To local disks,AND sends to another node for replication,a,vSphere,Write request IO flow,vSphere,b,c,Key facts:,Only reads are actually local writes MUST go to remote node,The read request can,only,come from one placewhat happens if the local CVM is busy?,Local IO,Remote IO,CONFIDENTIAL,Nutanix Data Locality:vMotion&Data Locality,Post vMotion Nutanix IO Flow:,VM has changed hosts so data temporarily not local,Reads:(,shown in green,),VM read request goes through CVM,Data not on local host send read request to former host,Cache read data to new local host,Eventually blocks will be written to new local node and data locality will be restored,Writes:(not shown),Essentially no change write to local host&second host,vSphere,Post vMotion read request IO flow,vSphere,New Host,Former Host,Key facts:,After vMotion initial reads are remote for some amount of time,The read request can,still only,come from one place,Remote IO,CONFIDENTIAL,VSAN Distributed Cache:Better Than Data Locality,VSAN IO Flow:,VM reads&writes are,distributed,to all hosts holding replicas,If one host is busy,the others can still service IO requests,vMotion,DRS,HA,etc.cause NO change to IO flow and no additional overhead,Key facts:,VSAN IO delivered from more than one location,10 GbE network overhead is small,The SDDC and SDS is about,abstraction,pooling,and,automation.,It should perform to,policy,specs,automatically,without worrying about placement of individual components.,VSAN delivers on this goal.,Distributed IO,Source:Duncan Epping.Sept 30,2013.,www.yellow-Details,性能测试:,Nutanix,的物理配置高于,VSAN,Hardware Details,VSAN(4 nodes),Nutanix(4 nodes),Server brand,Supermicro,Nutanix,Server model,SYS-2027TR-H70RF+,NX-3050,CPU model,Xeon E5-2650,Xeon E5-2670,CPU speed,2GHz,2.6GHz,Logical threads,32 total,32 total,RAM,128GB,256GB,SSD specs,Intel SSDSC2BB30 300GB,2 x Intel SSDSC2BA40 400GB(S3700),Storage controller,LSI2008,LSI SAS2008(passthrough mode),Benchmark Info,Benchmark name,DVD Store,URL,threads,0 think time,30 GB database,VM,configuration,Single VM with 4 vCPU,8 GB vRAM,61,Data Locality Benchmarking:Nutanix,Nutanix,两次,vMotion,后,性能下降很多,Nutanix depends on data locality:after vMotion data blocks must go from old CVM to new CVM,Orders per minute 5-minute moving average,vMotion,vMotion,Time(seconds),Orders per minute,Work
展开阅读全文