ExchangeServer白手起家系列系统SLA和监控流程

Slide Title,Body Text,Second Level,Third Level,Fourth Level,Fifth Level,Exchange Server,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,系列课程内容,万事开头难,进行企,业,业,IT,架构评,估,估,做一个,Exchange,管理员,必,必须掌,握,握的技,术,术,-,软件篇,做一个,Exchange,管理员,必,必须掌,握,握的技,术,术,-,硬件篇,邮件系,统,统设计,和,和选型,的,的原则,如何执,行,行压力,测,测试？,软件安,装,装和服,务,务器管,理,理的规,范,范化,用户和,系,系统迁,移,移过程,中,中的注,意,意事项,大规模,的,的用户,帐,帐号管,理,理、组,管,管理中,采,采用的,常,常见策,略,略,客户端,部,部署和,管,管理,邮件系,统,统日常,维,维护的,要,要点,系统,SLA,和监控,流,流程,邮件安,全,全、反,病,病毒和,反,反垃圾,邮,邮件,存储备,份,份策略,你的系,统,统可靠,性,性有几,个,个,9,？,Down,机了怎,么,么办？,常见问,题,题汇总,和,和答疑,今日议,程,程,SLA,是反映,一,一项,IT,服务的,综,综合指,标,标，不,仅,仅仅代,表,表系统,的,的在线,时,时间，,更,更反映,了,了用户,对,对系统,的,的认可,程,程度。,本,本次课,程,程主要,讨,讨论构,成,成,SLA,的组成,部,部分，,以,以及如,何,何通过,有,有效的,工,工作来,提,提升,SLA,和用户,的,的满意,程,程度。,SLA,的概念,和,和深入,解,解析,关键业,务,务系统,监,监控的,流,流程,Exchange,主要的,性,性能指,标,标,性能优,化,化建议,SLA,的概念,和,和深入,解,解析,SLA,：,Service Level Agreement,是在确,定,定的成,本,本下与,客,客户达,成,成的服,务,务和系,统,统使用,情,情况的,协,协议。,最简单,的,的指标,：,：系统,服,服务在,线,线时间,更进一,步,步的衡,量,量：,资源：,软,软硬件,的,的使用,效,效率和,投,投资回,报,报,性能：,用,用户的,使,使用体,验,验满意,程,程度,服务：,各,各类问,题,题和帮,助,助请求,的,的解决,速,速度,服务级,别,别协议,(SLA),的结构,Service Based-One ServiceLevelAgreement fora specific servicefor allcustomersofthatservice,CustomerBased-OneService Level Agreementforaspecificcustomerforall services,Multi-Level,CorporateLevel-coveringallgenericService Level Management(SLM)issues thatwouldbeappropriatefor allcustomers,Customer Level-covering allService Level Management(SLM)issues relevant to a specificcustomer group for allservices used,Service Level-coveringall ServiceLevelManagement(SLM)issuesrelevant toa specificservice fora specific customer group,如何制定,SLA,在项目实施阶,段,段的前期了解,用,用户对,SLA,的需求,在实施过程中,根,根据实际的软,硬,硬件投入情况,进,进行动态的更,新,新，并及时与,客,客户,(,用户,),进行沟通,如果项目涉及,到,到第三方供应,商,商，应在制定,SLA,之前取得第三,方,方的认可和确,认,认,如何贯彻,SLA,服务器端,服务器配置和,部,部署的规范统,一,一,运营过程中遵,循,循严格的流程,变更管理,最终用户,培训,Helpdesk,丰富客户端的,功,功能，增强用,户,户体验,整体协调,IT,支持部门内部,的,的横向沟通和,协,协调,管理与第三方,供,供应商的关系,及时保持与客,户,户,(IT,主管部门，,CIO,等,),的沟通,如何衡量,SLA,服务器在线时,间,间,客户满意度,服务器在线时,间,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,Mean Time to Repair(,MTTR,)“,总的停机时间,Time Between System Incidents,(,MTBSI,),Time Between Failures,(,MTBF,),客户满意度,客观的设定用,户,户的期望值,建立良好的,Helpdesk,流程,如果是跨国企,业,业，,Helpdesk,流程的一致性,非,非常重要,性能,可扩展性,可用性,关键业务系统,监,监控的流程,每日检查的内,容,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,重要变更,每日检查报表,的,的分发,建议使用,SharePoint,网站保存每日,检,检查的报表,学会使用性能,计,计数器来给,Exchange,把脉,demo,性能优化建议,Smalldesignchanges canreapbig benefits,Do notbe overly aggressive when tuning,especially when itinvolves registrychanges,Alwaysuse the MicrosoftExchange Server Best Practices Analyzer Tool for validation,Storage Performance,The biggestcauseof Exchangeperformanceissues,Usually toofew spindlesto meet thedemand,Exchange isI/O hungry,Back of theenvelope:1spindle forevery100 users,Actions to take,Work out your I/Oprofile,Tell your storagevendorhow many IOPS youwillproduce,Set cachingcontrollersto 100%write/4,kilobyte(,KB)cache page size,Thinkend-to-end storage(forexample,Host BusAdaptors),Use DiskParto align tracks(performance boost),Use JetStress to validate yourstorage architecture,Read the paper,Optimizing Storagein Exchange2003,!Warning Signs!,Average disklatency isgreater than20,milliseconds,Long disk latencyspikesare greaterthan50 milliseconds,Fast Accessto Active Directory,If Exchangeis slow accessingthe Active Directory,directory service,store performance suffers and maildelivery isslow,Doublecheckthatthe,Directory ServiceAccess(,DSAccess)list only contains local domaincontroller or global catalog servers,Remember that roundrobinis used,Round-triptimes should be less than 10ms,Ruleof thumb,You shouldhaveoneglobal,catalog,processorfor everyfourExchange processors,Verify that the/3GBswitch isbeing used onglobal,catalogs,IncreasesJET cachefrom512 megabytes(MB)to 1 gigabyte(GB),Moreof the Active Directory is cached in memory,Correct Use of/3GBSwitch,Mailbox and public folderservers,Implementwhenthereis=1 GB RAM,Windows Server2003implement/USERVA=3030,Global catalogservers,Implementwhenthereis=1 GB RAM,Front-endand routing servers,Do not implement regardless ofmemory,Instead,tune memoryforsystem cache,/3GBand/USERVA Switches,UserMode,Kernel Mode,User Mode,User Mode,Kernel Mode,User Mode,HeapDeCommitFreeBlockThreshold,Should beset to 0 x40000on all Exchangeservers,Reduces virtualmemory fragmentation bycoalescingsmall free memory regions into larger blocks,EspeciallyimportantforlargeExchangemailbox servers,9582 begone!,http:/ Hyper-Threading,Setting inBIOS,Sometimesdefaults to OFF,Can improve Exchangeprocessorefficiency by25%,WorkswithWindows Server2003andWindows 2000 Server,Deploy ExchangeServer 2003 SP2,30%reduction in I/Opattern,Drastic reduction inlogfilereplay times,From30 secondsto less than 2seconds,Support for 8-way hyper-threading servers,SetsMaxDSNSizeto 10 MBby default,Side:Detectionof single-bitfliperrors,Hardware DeviceDrivers,NonPaged Pool and page tableentry(PTE)efficiency,Disable unusedhardware devices,Especiallyconsidernetwork adapters andstorage cards,Use identical hardware insidethe server,One driverformultiple devices,Be especi