ExchangeServer白手起家系列系统SLA和监控流程

白手起家系列系统SLA和监控流程喻勇,PMP/MCSE微软最有价值专家，特约讲师传世系列课程内容万事开头难进行企业IT架构评估做一个Exchange管理员必须掌握的技术-软件篇做一个Exchange管理员必须掌握的技术-硬件篇邮件系统设计和选型的原则如何执行压力测试？软件安装和服务器管理的规范化用户和系统迁移过程中的注意事项大规模的用户帐号管理、组管理中采用的常见策略客户端部署和管理邮件系统日常维护的要点系统SLA和监控流程邮件安全、反病毒和反垃圾邮件存储备份策略你的系统可靠性有几个9？Down机了怎么办？常见问题汇总和答疑今日议程SLA是反映一项IT服务的综合指标，不仅仅代表系统的在线时间，更反映了用户对系统的认可程度。本次课程主要讨论构成SLA的组成部分，以及如何通过有效的工作来提升SLA和用户的满意程度。SLA的概念和深入解析关键业务系统监控的流程Exchange主要的性能指标性能优化建议的概念和深入解析SLA：ServiceLevelAgreement是在确定的成本下与客户达成的服务和系统使用情况的协议。最简单的指标：系统服务在线时间更进一步的衡量：资源：软硬件的使用效率和投资回报性能：用户的使用体验满意程度服务：各类问题和帮助请求的解决速度服务级别协议(SLA)的结构ServiceBased-OneServiceLevelAgreementforaspecificserviceforallcustomersofthatserviceCustomerBased-OneServiceLevelAgreementforaspecificcustomerforallservicesMulti-LevelCorporateLevel-coveringallgenericServiceLevelManagement(SLM)issuesthatwouldbeappropriateforallcustomersCustomerLevel-coveringallServiceLevelManagement(SLM)issuesrelevanttoaspecificcustomergroupforallservicesusedServiceLevel-coveringallServiceLevelManagement(SLM)issuesrelevanttoaspecificserviceforaspecificcustomergroup如何制定SLA在项目实施阶段的前期了解用户对SLA的需求在实施过程中根据实际的软硬件投入情况进行动态的更新，并及时与客户(用户)进行沟通如果项目涉及到第三方供应商，应在制定SLA之前取得第三方的认可和确认如何贯彻SLA服务器端服务器配置和部署的规范统一运营过程中遵循严格的流程变更管理最终用户培训Helpdesk丰富客户端的功能，增强用户体验整体协调IT支持部门内部的横向沟通和协调管理与第三方供应商的关系及时保持与客户(IT主管部门，CIO等)的沟通如何衡量SLA服务器在线时间客户满意度服务器在线时间故障发生检测到问题检测时间诊断响应时间修复恢复服务问题解决下一次故障修复时间恢复时间MeanTimetoRepair(MTTR)“总的停机时间TimeBetweenSystemIncidents(MTBSI)TimeBetweenFailures(MTBF)客户满意度客观的设定用户的期望值建立良好的Helpdesk流程如果是跨国企业，Helpdesk流程的一致性非常重要性能可扩展性可用性关键业务系统监控的流程每日检查的内容备份情况邮件队列系统性能磁盘空间硬件情况昨天或昨晚的重要变更每日检查报表的分发建议使用SharePoint网站保存每日检查的报表学会使用性能计数器来给Exchange把脉demo性能优化建议SmalldesignchangescanreapbigbenefitsDonotbeoverlyaggressivewhentuning,especiallywhenitinvolvesregistrychangesAlwaysusetheMicrosoftExchangeServerBestPracticesAnalyzerToolforvalidationThebiggestcauseofExchangeperformanceissuesUsuallytoofewspindlestomeetthedemandExchangeisI/OhungryBackoftheenvelope:1spindleforevery100usersActionstotakeWorkoutyourI/OprofileTellyourstoragevendorhowmanyIOPSyouwillproduceSetcachingcontrollersto100%write/4kilobyte(KB)cachepagesizeThinkend-to-endstorage(forexample,HostBusAdaptors)UseDiskPartoaligntracks(performanceboost)UseJetStresstovalidateyourstoragearchitectureReadthepaper,OptimizingStorageinExchange2003!WarningSigns!Averagedisklatencyisgreaterthan20millisecondsLongdisklatencyspikesaregreaterthan50millisecondsIfExchangeisslowaccessingtheActiveDirectorydirectoryservice,storeperformancesuffersandmaildeliveryisslowDoublecheckthattheDirectoryServiceAccess(DSAccess)listonlycontainslocaldomaincontrollerorglobalcatalogserversRememberthatroundrobinisusedRound-triptimesshouldbelessthan10msRuleofthumbYoushouldhaveoneglobalcatalogprocessorforeveryfourExchangeprocessorsVerifythatthe/3GBswitchisbeingusedonglobalcatalogsIncreasesJETcachefrom512megabytes(MB)to1gigabyte(GB)MoreoftheActiveDirectoryiscachedinmemoryMailboxandpublicfolderserversImplementwhenthereis=1GBRAMWindowsServer2003implement/USERVA=3030GlobalcatalogserversImplementwhenthereis=1GBRAMFront-endandroutingserversDonotimplementregardlessofmemoryInstead,tunememoryforsystemcache://Shouldbesetto0x40000onallExchangeserversReducesvirtualmemoryfragmentationbycoalescingsmallfreememoryregionsintolargerblocksEspeciallyimportantforlargeExchangemailboxservers–9582begone!=315407SettinginBIOSSometimesdefaultstoOFFCanimproveExchangeprocessorefficiencyby25%WorkswithWindowsServer2003andWindows2000Server30%reductioninI/OpatternDrasticreductioninlogfilereplaytimesFrom30secondstolessthan2secondsSupportfor8-wayhyper-threadingserversSetsMaxDSNSizeto10MBbydefaultSide:Detectionofsingle-bitfliperrorsNonPagedPoolandpagetableentry(PTE)efficiencyDisableunusedhardwaredevicesEspeciallyconsidernetworkadaptersandstoragecardsUseidenticalhardwareinsidetheserverOnedriverf