在虛擬化的上運(yùn)行關(guān)鍵應(yīng)用U2VL

VMwareInc.在虛擬化的平臺上運(yùn)轉(zhuǎn)

關(guān)鍵運(yùn)用

U2VL內(nèi)容提要什么是關(guān)鍵運(yùn)用關(guān)鍵運(yùn)用為何虛擬化關(guān)鍵運(yùn)用如何虛擬化結(jié)語什么是關(guān)鍵運(yùn)用?數(shù)據(jù)庫效力器Web效力器運(yùn)用效力器42%47%2%對關(guān)鍵運(yùn)用而言,VMware是DisruptiveTechnology100%98%59%DB2DB2Oracle71國內(nèi)案例

中國人壽股份國家電網(wǎng)公司中國南方電網(wǎng)北京市朝陽區(qū)教育委員會蘇州工業(yè)園區(qū)管委會計算機(jī)信息中心廣東中煙山東電信江西財政廳遵義商行西南交通大學(xué)圖書館合肥卷煙廠：精品黃山消費(fèi)系統(tǒng)，OracleRAC做數(shù)據(jù)庫高可用，底層存儲是IBMSVC＋DS5000山東城商聯(lián)盟：銀行基金購買運(yùn)用，VMwareHA做系統(tǒng)高可用,四川電信：綜合營帳的計費(fèi)，VMwareHA。信譽(yù)控制,OracleRAC江蘇電信:網(wǎng)上營業(yè)廳ORACLERAC這些堅決的從Unix移植到Vmware-Linux的企業(yè)都同時看到了什麼？性能VMware+x86Linux與RS/6000-AIX性能一樣好。風(fēng)險VMware+x86Linux比RS/6000-AIX更可靠。處理了UNIX老問題VMware+x86Linux處理了困擾RS/6000-AIX的四大IT難題。Agenda什么是關(guān)鍵運(yùn)用關(guān)鍵運(yùn)用為何虛擬化關(guān)鍵運(yùn)用如何虛擬化結(jié)語更穩(wěn)定的性能，更可靠處理四大IT難題X86-E7-ESX性能優(yōu)于Power621%銀行基金WAS+DB2運(yùn)用性能測試比較x86與Power6關(guān)鍵運(yùn)用虛擬化性能比RS/6000好嗎？可靠嗎？XX保險公司OracleOLTP在線買賣–虛擬化最終測試結(jié)果結(jié)論：x86Xeon-7650比Power6CPU處置才干高33.7%慘HPBL685G7，64Core目前配置的功耗約為600W。IBMP6/57016core配置的功耗為3200W踏破鐵鞋：X86效力器與Power7小機(jī)TPC-C的比較每個核的性能Power7優(yōu)于x8616.8%每個核的性能Power7比虛擬化后的x86快12%Vmware數(shù)據(jù)庫虛擬化后的損耗為5.8%SAP-2tier比較關(guān)鍵運(yùn)用x86虛擬化與Power7比較VMware+x86Linux與PowerVM-AIX的性能比較X86-E7-ESX性能優(yōu)于Power621%X86-7560-ESX性能優(yōu)于Power633.7%X86-7560-ESX+SSD性能優(yōu)于Power637%基于TPC-CPower7性能優(yōu)于x8616.8%基于SAPPower7性能優(yōu)于x86+VMware12%X86+VMware性能大于Power6+PowerVM30%Power7+PowerVM性能大于x86+VMware15%Power7++PowerVM性能大于x86+VMware30%性能我清楚了，不用再測了。接著我們看HA的測試結(jié)果，假設(shè)可以，我們可以開場規(guī)劃一步一步的把小機(jī)換成x86采用新技術(shù)要有遠(yuǎn)見，2021年我開場用了VMware以后，2年我?guī)凸臼×?000萬效力器的采購款。這還不包含R6的MA跟軟件的維護(hù)費(fèi)用。X86加了VMware以后的可靠性超越AIX加上PowerVMware是大才小用，它真正的價值在可靠性，不是省那些CPU.那些x86的cpu沒幾個錢Stephen:程總，您是說VMware加上x86可靠性超越小機(jī)嗎？程總：是啊，遠(yuǎn)遠(yuǎn)超越。我們過去幾年出的事故都是小機(jī)，幾乎沒有x86。AIX數(shù)據(jù)庫PowerVMHypervisorHA-CMPPwrHA運(yùn)用Novell等等ESXHA數(shù)據(jù)庫運(yùn)用AIX數(shù)據(jù)庫PowerVMHypervisorHA-CMPPwrHA運(yùn)用Novell等等ESXHA數(shù)據(jù)庫運(yùn)用一樣的硬件，但是：x86虛擬化平臺提供更好的可用性AIXAIX要不斷的堅持一致性SUSE效力器的可靠性依賴的是高可用的軟件IT管理員：由于指點(diǎn)要求零事故啊！指點(diǎn)：我們RS/6000從來都沒有事故啊！Step1.InstallthePowerHAsoftwareYoucanpurchasethissoftwarefromIBMdirectly(seeResourcesforalink);thefilesetsallstartwiththewordcluster.Usetheinstallpcommandtoinstallthesoftware,muchlikeanyotherlicensedprogrampackage(LPP).Step2.EditsomeflatfilesPutalloftheIPaddressesassociatedwiththecluster—boot,persistent,andservice—intoeach/etc/hostsfileoneachnodeofthecluster.Dothesamewiththe/usr/es/sbin/cluster/etc/rhostsfile.VerifythattheserverhostnamesmatchtheappropriateIPaddresses;theserver'shostnameshouldalsomatchwiththepersistentIPaddress.Step3.ConfigurethebootIPaddressesRunthesmittychinetcommand,andsetthebootIPaddressesforeachnetworkadapter.Makesurethatyouareabletopingandconnectfreelyfromnodetonodeonallrespectivenetworks.Also,double-checktomakesurethatthedefaultrouteisproperlyconfigured.Ifitisn't,runsmittytcpip,gointotheMinimumConfigurationmenu,enterthedefaultroutefortheprimaryadapter,andpressEnter.Step4.MakeapplicationstartandstopscriptsCreatetwosimpleKornshellscripts—onethatstartsanapplicationandonethatstopsanapplication.Keepthesescriptsinidenticaldirectoriesonbothnodes.Step5.DefinetheclusterRunthecommand:smittycm_config_an_hacmp_cluster_menu_dmn

Then,definethecluster,includingnamingitappropriately.Step6.DefinethenodesRunthecommand:smittycm_config_hacmp_nodes_menu_dmn

Defineeverynodewithintheclusteronbothnodes.Step7.DefinethenetworksRunthecommand:smittycm_config_hacmp_networks_menu_dmn

ThisdefinesonenetworkperEthernetadapter.IprefertousethePre-definedoptionasopposedtotheDiscoveredpath,butthatisuptoyourdiscretion.Checkthesubnetmasksforconsistency.Step8.DefinethebootIPaddressesRunthecommand:smittycm_config_hacmp_communication_interfaces_devices_menu_dmn

ThisdefinesthebootIPaddressesontherespectivenetworkadapters.ThisaddressshouldbethesameIPaddressesyouusedinstep3.MakesureyoudefinetheseaddresseswithintheproperrespectivePowerHA-definednetwork.Step9.DefinethepersistentIPaddressesRunthecommand:smittycm_config_hacmp_persistent_node_ip_label_addresses_menu_dmn

ThisdefinesthepersistentIPaddresses,againpayingattentiontopicktheproperrespectivePowerHA-definednetwork.Step10.DefinetheserviceIPaddressesRunthecommand:smittycm_config_hacmp_service_ip_labels_addresses_menu_dmn

ThisdefinestheserviceIPaddresses.Step11.PerformadiscoveryandrebootBythispoint,thenodesshouldhavetheabilitytocommunicatewitheachotherandkeeptheinformationstoredinthenodes'ObjectDataManagers(ODMs)insync.Makethenodeswithintheclustercommunicatewitheachotherbyrunningthecommand:smittycm_extended_config_menu_dmn

SelecttheDiscoverPowerHA-relatedInformationfromConfiguredNodesoption,andcheckforerrorstofix.Generally,rebootingeachnodecanclearupanyminorproblems,andthisisagoodpointtotestrestartingeachserveranyway.Step12.DefinetheresourcegroupRunthecommand:smittycm_hacmp_extended_resource_group_config_menu_dmn

Definetheresourcegroup.Then,performthesesteps:Selectallparticipatingnodesinthecluster.SettheStartupPolicytoOnlineOnFirstAvailableNode.SettheFallbackPolicytoNeverFallback.Thissettingpreventstheresourcesfromgoingbacktotheoriginalserverwhenitisbroughtup,whichisawisethingtodo.Step13.CreateasharedvolumegroupNote:Runthiscommandonlyononenode.Runthesmittycl_vgcommand,andcreateasharedvolumegroup.Whenyoucreateasharedvolumegroup,youonlyneedtoselectoneofthenodes,becausethediskisshared.Step14.CreateaheartbeatdiskFirst,runthesmittycl_convgcommand,andthenselectCreateaConcurrentVolumeGroupwithDataPathDevices.Chooseonenodeandthetargetdisk.Next,runthecommand:smittycm_config_hacmp_communication_interfaces_devices_menu_dmn

Repeatstep7,exceptthistime,selecttheDiscoveredoptionandthetargetdisk.Step15.DefineanapplicationserverRunthecommand:smittycm_cfg_app_extended

ThisdefinesanapplicationserverforanapplicationthatPowerHAwillmanage.Usethescriptsyoucreatedinstep4.Step16.ConfiguretheresourcegroupNote:Runthiscommandononlyonenode.Runthecommand:smittycm_hacmp_extended_resource_group_config_menu_dmn

SelecttheChange/ShowResourcesandAttributesforaResourceGroupoption.Then,performthesesteps:SelecttheappropriateserviceIPaddresses.Selecttheappropriatesharedvolumegroupsandheartbeatdisk.Selecttheappropriateapplicationservers.Step17.PerformasynchronizationSynchronizetheclusterconfiguration.Runthecommand:smittycm_ver_and_sync

SetAutomaticallycorrecterrorsfoundduringverification?toInteractive.Correctanyproblemsalongtheway.Step18.StarttheclusterAtthispoint,theclusterisreadytostart.Ononeofthenodes,runthesmittyclstartcommand,andpickthatparticularnode.Mypreferenceisnottohavetheclusterstartonreboot,becauseifthereisaPowerHA-relatedproblemonstartup,itcanbedifficulttotroubleshootit.Afterthenodecomesupwiththeresourcesavailable,starttheclusterontheothernode.Step19.PerformatestfailoverThebestwayIhavefoundtotestPowerHA’sadaptabilityistoreboottheactivenodeandletthingsfailovernaturallywhilerunningthetail–f/tmp/hacmp.outcommandontheothernodetowatchasthingsgoover.Or,runthecommand:smittycl_resgrp_move.node_site

SelecttheMoveResourceGroupstoAnotherNodeoption.Step20.PerformfailuretestingIfyoureallywanttomakesureyourclusterissolid,performtestingbyliterallyemovingcablesandseeinghowtheresourcesmovebackandforth.Themoreyoutest,themorereliableyourclusterwillbe.Step1.InstallthePowerHA

Step2.Editsomeflatfiles

PutalloftheIPaddressesassociatedwiththecluster—boot,persistent,andservice—intoeach/etc/hostsfileoneachnodeofthecluster.Dothesamewiththe/usr/es/sbin/cluster/etc/rhostsfile.VerifythattheserverhostnamesmatchtheappropriateIPaddresses;theserver‘shostnameshouldalsomatchwiththepersistentIPaddress.

Step3.ConfigurethebootIPaddresses

Runthesmittychinetcommand,andsetthebootIPaddressesforeachnetworkadapter.Makesurethatyouareabletopingandconnectfreelyfromnodetonodeonallrespectivenetworks.Also,double-checktomakesurethatthedefaultrouteisproperlyconfigured.Ifitisn’t,runsmittytcpip,gointotheMinimumConfigurationmenu,enterthedefaultroutefortheprimaryadapter,andpressEnter.

Step4.Makeapplicationstartandstopscripts

CreatetwosimpleKornshellscripts—onethatstartsanapplicationandonethatstopsanapplication.Keepthesescriptsinidenticaldirectoriesonbothnodes.、VmwareHA-保證關(guān)鍵運(yùn)用VMware處理了AIX的四大IT難題！沒有容錯功能沒有系統(tǒng)飄移功能〔VMotion〕沒有基于PowerVM的災(zāi)備處理方案電力緊缺超低運(yùn)用率不支持開源軟件HACMP/PowerHA的可靠性問題被單一廠商綁架oracle@u2lu2l]$pythonu2l.pyEnternumberoflooptotest->50transaction#0FriSep2115:01:382021response=1transaction#1FriSep2115:01:392021response=1transaction#2FriSep2115:01:412021response=1transaction#3FriSep2115:01:422021response=1transaction#4FriSep2115:01:432021response=1Sleep1second&trytoaccessagain...Sleep2second&trytoaccessagain...Sleep3second&trytoaccessagain...Sleep4second&trytoaccessagain...。。Sleep130second&trytoaccessagain...Sleep131second&trytoaccessagain...Sleep132second&trytoaccessagain...transaction#23FriSep2115:02:342021response=133transaction#24FriSep2115:02:362021response=1transaction#25FriSep2115:02:372021response=1。。transaction#30FriSep2115:02:442021response=1transaction#31FriSep2115:02:452021response=1transaction#32FriSep2115:02:472021response=1Testfor50loopendat->FriSep2115:03:102021運(yùn)用Vmware-HA高可用接納oracle@u2lu2l]$pythonu2l.pyEnternumberoflooptotest->50transaction#0FriSep2115:01:382021response=1transaction#1FriSep2115:01:392021response=1transaction#2FriSep2115:01:412021response=1transaction#3FriSep2115:01:422021response=1transaction#4FriSep2115:01:432021response=1transaction#1FriSep2115:01:392021response=1transaction#2FriSep2115:01:412021response=1transaction#3FriSep2115:01:422021response=1transaction#4FriSep2115:01:432021response=1transaction#1FriSep2115:01:392021response=1transaction#2FriSep2115:01:412021response=1transaction#3FriSep2115:01:422021response=1transaction#4FriSep2115:01:432021response=1transaction#1FriSep2115:01:392021response=1transaction#2FriSep2115:01:412021response=1transaction#3FriSep2115:01:422021response=1transaction#4FriSep2115:01:432021response=1。。。transaction#30FriSep2115:02:442021response=1transaction#31FriSep2115:02:452021response=1transaction#32FriSep2115:02:472021response=1Testfor50loopendat->FriSep2115:03:102021運(yùn)用Vmware-FT容錯接納133-140秒0秒AIX-PowerVM的第一個難題：沒有容錯功能RS/6000p6-595關(guān)鍵運(yùn)用2021年5月運(yùn)用形狀I(lǐng)BMp6-P595，數(shù)據(jù)效力器，AIX5.3，4.2GHz，2x4=8core.第四年用戶為何不情愿提高運(yùn)用率？RS/6000運(yùn)用率低為何是常見景象？用戶：噢，我們也有這個景象，好似挺普遍的。第四年Server90%以上的破費(fèi)Oracle90%以上的破費(fèi)ServerMA-90%以上的破費(fèi)OracleMA-90%以上的破費(fèi)數(shù)據(jù)庫效力器Web效力器運(yùn)用效力器42%47%2%對關(guān)鍵運(yùn)用而言,VMware是DisruptiveTechnologyPowerVMware+AIXVMware+LinuxAIX-PowerVM的第二個難題沒有系統(tǒng)飄移功能〔VMotion〕特點(diǎn)：一舉數(shù)得-可靠，易于實現(xiàn)異步，不實時先保證數(shù)據(jù)逐漸遷移運(yùn)用發(fā)生災(zāi)難的保證AIX-PowerVM的第三個難題：沒有基于PowerVM的災(zāi)備處理方案消費(fèi)中心VirtualCenterSiteRecoveryManagerSiteADatastoreGroupsIBM-x3850HP-DL980災(zāi)備中心VirtualCenterSiteRecoveryManagerSiteBDatastoreGroupsIBM-x3850HP-DL980ReplicationHACMP/PowerHAHMC/PowerVMSiteA/RS6000R6-p780R6-p795REDOLOGReplication云災(zāi)備：一舉多得–5贏運(yùn)轉(zhuǎn)商：電信效力器設(shè)備供應(yīng)商軟件與技術(shù)：VMware效力提供商：協(xié)作同伴需求災(zāi)備的中小企業(yè)theregister.co.uk/2021/08/31/ibm_power_chip_roadmap_update/電不夠用，拿掉一臺PowerVM-AIX機(jī)可以低三臺AIX-PowerVM的第四個難題：耗電AIX-PowerVM的第五個難題運(yùn)轉(zhuǎn)舊版本的AIX極其困難Agenda什么是關(guān)鍵運(yùn)用關(guān)鍵運(yùn)用為何虛擬化關(guān)鍵運(yùn)用如何虛擬化結(jié)語當(dāng)前運(yùn)轉(zhuǎn)數(shù)據(jù)庫關(guān)鍵運(yùn)用最可靠的效力器架構(gòu)是Vmware搭建的Linux效力器，不是PowerVM下搭建的AIX