云计算技术与架构前瞻-张福波.ppt

Copyright 2010 Platform Computing Corporation.All Rights Reserved.,*,Proprietary and Confidential:Platform Computing Corporation,Copyright 2010 Platform Computing Corporation.All Rights Reserved.,Copyright 2009 Platform Computing Corporation.All Rights Reserved.,Copyright 2009 Platform Computing Corporation.All Rights Reserved.,Copyright 2010 Platform Computing Corporation.All Rights Reserved.,*,*,*,Proprietary and Confidential:Platform Computing Corporation,Copyright 2010 Platform Computing Corporation.All Rights Reserved.,*,*,*,Proprietary and Confidential:Platform Computing Corporation,Copyright 2010 Platform Computing Corporation.All Rights Reserved.,*,*,*,Proprietary and Confidential:Platform Computing Corporation,Copyright 2010 Platform Computing Corporation.All Rights Reserved.,*,*,*,Proprietary and Confidential:Platform Computing Corporation,Copyright 2010 Platform Computing Corporation.All Rights Reserved.,*,*,云计算技术与架构前瞻,Presented by:Fubo Zhang,云计算的市场前景,云计算的现状,云计算的技术、架构发展趋势,内容提要,3,1979:CFD Contributions to the 767,High-Speed Wing,Design,Cab Design,Engine/Airframe Integration,Inlet Design,Wing-Body,Fairing Design,Nacelle Design,Much CFD penetration,.,Opportunities exist for higher,accuracy and expanded complexity,Some CFD penetration,.,Opportunities exist for large increases in design process speed and application,CFD penetration opportunity,4,2005:CFD Contributions to the 787,High-Speed Wing,Design,Cab Design,Engine/Airframe Integration,Inlet Design,Inlet Certification,Exhaust,System Design,Cabin,Noise,Community Noise,Wing-Body,Fairing Design,Vertical Tail and Aft Body Design,Design For,Stability&,Control,High-Lift Wing Design,APU Inlet,And Ducting,ECS Inlet Design,APU and Propulsion,Fire Suppression,Nacelle Design,Thrust Reverser,Design,Design for FOD,Prevention,Aeroelastics,Much CFD penetration,.,Opportunities exist for higher,accuracy and expanded complexity,Some CFD penetration,.,Opportunities exist for large increases in design process speed and application,Icing,Air Data System Location,Connexion Antenna,Vortex Generator Placement,Planform Design,Buffet Boundary,Wake Vortex Alleviation,Reynolds Number Corrections for Loads and S&C,Flutter,Control Failure Analysis,Wind Tunnel Corrections,Design For,Loads,Wing Tip Design,Wing Controls,Avionics Cooling,Interior Air Quality,Engine Bay Thermal Analysis,CFD penetration opportunity,Gartner View of Private Cloud Market,可升级的,共享,自动化,弹性计算,451 Group View of the Cloud Market,管理,自动化,安全,存储,Forrester View on Cloud Computing Platforms,云计算实现的,3,条路径,Time,Scope of sharing,2003,2015,2009,Client/Server,Silos,Workgroup,/LOB,VM Cluster,同构服务器的整合,有限的应用,基于,VM,的管理功能,网格,静态资源、动态调度,基于策略的调度,Enterprise,Public&Private Cloud,9,12/17/2024,9,公共云,-What do we believe?,ESX,Virtual Machine,Virtual Machine,Virtual Machine,Virtual Machine,Virtual Machine,Virtual Machine,物理机,PC cloud,手机,移动上网,3G,上网机,Internet,接入上网,Workstation,VIP,接入,10,虚拟机,12/17/2024,10,公共云,-What do I believe?,11,PC cloud,上网笔记本等终端设备,台式电脑、笔记本将会被淘汰,虚拟,PC,服务运营商将会替代,MS,和,google,Data Bank,解决数据、信息安全的最终解决方案,衍生许多种数据服务，比如医疗健康，个人企业信用征询，个人数据银行,Many SPs,PC SP,Storage SP,Game SP,Many SAAS SPs,Many other SPs,12/17/2024,11,Platform ISF,HPC Apps,Test&Dev,Analytics,LSF,Symphony,LSF,Symphony,Private cloud management software,私有云,Java,JBoss,Tomcat,12/17/2024,12,数据中心体系很复杂，系统的维护和管理难度大,IT,成本高，资源占用多，配置峰值资源需求,=,浪费资源,系统稳定性、可靠性低。人工服务为主，高成本、低满意度,IT,传统模式不能适应业务部署速度的需求（至少,2,个月）,地理上分布的用户,根据应用静态分配,IT,资源,Supply and Demand Not Aligned,Web,eCommerce,Intranet,ERP,CRM,BI,Dev/QA,Email,今天,IDC,数据中心的挑战,12/17/2024,13,云计算（虚拟化）带来的好处,资源虚拟化，从而实现全局共享,节省投资,SOA,框架使得,IT,能对业务需求提供快速部署，时间从原来的几个月缩短到半天,提高了业务支撑能力,一体化管理打破了原来,IT,烟囱式业务系统管理，使得系统维护人员的效率大大提高 运维效率提高,消除单点故障造成的稳定、可靠性问题 提高系统稳定性和可靠性,帮助,IT,投资风险和决策,Smart IT,节能减排 绿色,IT,12/17/2024,14,实现云计算（虚拟化）的困难,没有成熟的产品和方案让运营商有直观的理解。,各个厂商都在推、强调自己的云计算概念和产品，像盲人摸象，无法让用户有个全局的认识。,现有,IT,系统的改动成本和风险,传统业务系统虚拟化改造,少有成功案例可以借鉴,组织管理和人们的认识需要提高,12/17/2024,15,Platform,在实现云计算方面的优势,Platform,已经有了自己的云计算产品,ISF-IAAS,Platform,的云计算产品是建立在许多用户的需求上的，如,JP Morgen,CityBank,Amazon,Alatum,我们帮助用户实现小步快跑的策略，降低,IT,系统的改动成本和风险,我们本地有近,200,名研发团队来协助用户的业务系统虚拟化改造,Platform,已有几个云计算成功案例,Platform,可提供业务培训和成功用户案例考察,12/17/2024,16,Java Cloud:Fortune 500 FS Companies,App Server,App 1,App Server,Platform ISF,烟囱式的业务系统导致了系统部署时间长，资源利用率低下，,IT,管理维护成本高,Virtual Servers,Self-Service Portal,100s Java,应用软件运行在,1000s UNIX SMP,服务器,App Server,App 2,App Server,App N,App 1,App 2,App N,挑战,资源池方案：,Linux,Xen/KVM,和,storage;,自服务管理,调度自动化，管理一体化,系统部署的时间从几天缩短到几分钟,没有双机备份,低成本,解决方案,Benefits,环境,12/17/2024,17,SaaS Cloud:Fetch Technologies,Platform ISF,CUST A,CUST B,CUST C,CUST X,由于手动部署业务系统，无法满足日益增长的业务需求,简单的用户系统配置的改变，需要几个小时乃至几天,200 virtual servers in production and test/dev using VMware,利用私有云的技术,Platform ISF,自动管理和部署系统,满足了客户业务扩张的需求,减少了认为的中间环节，提高了工作效率,实现了,Service on Demand,解决方案,Benefits,环境,挑战,12/17/2024,18,IaaS Cloud:SingTel,在人工操作的方式下，管理和维护出租的计算机、计算用户使用情况和计费非常的不方便,Xen,VMware,Hyper-V,and physical blades on 2,400 cores providing public cloud services,挑战,利用私有云的技术,Platform ISF,自动管理和部署系统,多快好省的为用户提供各种服务,通过自动化管理，降低了维护费用,提升了市场竞争力,解决方案,Benefits,环境,12/17/2024,19,HPC Cloud:CERN,系统无法为用户提供灵活的资源配置，需要等待很长的时间,RHEL,KVM,Xen,Hyper-V for widely diverse set of scientific applications on x86 commodity boxes,挑战,利用私有云的技术,Platform ISF,自动管理和部署系统,在固有的资源上，开发了更多的服务项目,;,对一些应用系统的运行效率提高了一倍多,降低成本，提高资源利用率,通过自动化管理，降低了维护费用,解决方案,Benefits,环境,User B,Workload,Manager,Workload,Manager,User A,User A,User B,Platform LSF,Platform ISF,12/17/2024,20,3,个层次的云计算服务2024/12/17,21,Each level provides a higher level of service abstraction&may use lower level(s),服务器，网络，存储按需分配,e.g.,Amazon EC2,S3,IaaS,(,Infrastructure as a Service,，基础架构即服务,),APIs for developing new Cloud apps,e.g.,Google App Web,PaaS,(,Platform as a Service,，平台即服务,),应用软件对最终用户按需分配,e.g.,S,SaaS,(,Software as a Service,，软件即服务,),12/17/2024,21,云计算体系架构,服务调度和管理引擎,云计算平台,VM,中间构建,IAAS,调度分配引擎,资源池,IAAS,服务管理,SAAS,PAAS,服务系统,IAAS,服务平台,服务应用系统,云计算门户,IDC,服务,12/17/2024,22,Placement Engine Policy,12/17/2024,23,Placement,Engine,1)Application,Topology,DB,App,Server,App,Server,App,Server,Web,Server,Web,Server,2)Resource,Hierarchy,3)Network Access Cost Matrix(Available Bandwidth/Dynamic Utilization),DC,Rack,Group,Rack,Group,Rack,Group,Rack,Rack,Tier 1,Tier 2,Tier 3,Rack,Rack,Rack,Rack,DC1,DC2,DC3,DC1,DC2,DC3,Rack1,Rack2,Rack3,R2,R2,R3,Output:,Initial Placement,-Run-Time Management,-Defragmentations&Migration,Offering 1,300,Offering 2,200,Offering 3,400,Jan,Dec,Contract#123,Contract#444,Contract,#677,Contract#677,Contract#888,Contract#888,#999,Contract#78934,Contract#768689,Contract,#55,Contract,#889,DC,/,4)Previous reservations committed&Account Quotas,5)C-language plug-in placement policy:,vem_policy_init(),vem_policy_fin(),vem_policy_distribute(),vem_policy_place(),vem_policy_housekeep(),vem_policy_status(),vem_policy_info(),12/17/2024,23,基于资源可用状态的分配策略,Packing,（填满）,应用程序在尽量少的物理服务器上运行,每个被使用的服务器利用率最大化，减少资源碎片，减少能源消耗。,Striping,（分散）,应用程序分散在尽可能多的物理服务器上运行,减少机器故障带来的影响，提高应用程序的运行效率,Load-Aware,（基于负载）,新的应用程序总是运行在负载最轻的物理服务器上,获得更高的应用程序运行效率,HA-Aware,（基于高可靠性配置）,将支持高可靠性（,HA,）的服务器资源分配给关键业务,提供更高的资源可用性,Energy-Aware,（节能）,根据节能指数和数据中心热点运行应用程序,减少能源消耗,Affinity-Aware,（基于关联程度）,将任务分配到与关键资源关联度最高的服务器上，例如将任务分配到与存储系统直连的服务器上,保证应用程序运行效率,Server Model-Aware,（基于服务器类型）,根据服务器类型分配资源,尽可能利用性能好的、昂贵的资源，是投资回报最大化,Topology-Aware,（基于网络拓扑）,尽量使用连接到同一个交换机、背板、刀片中心的服务器运行用户任务,提高应用程序运行效率,12/17/2024,24,PTM Framework,Application-Aware,调度,ISF Master,Application,Orchestrator,KVM Hypervisor,Resource Group,Deploy&,Monitor VMs,Monitor,Get historical,usage,Grow/Shrink,Vmware Hypervisor,Resource Group,Physical,Linux Resource Group,Application Performance,Manager,ICM,PVMO,PTM,Call ISF or send,Alert,1,3,2,需求,:,动态的调整业务应用对资源的需求改变，而无须修改应用的中间件,输入：,系统负载参数,(cpu,mem),集成的应用软件效率管理系统,需求的调度策略,12/17/2024,25,Add custom attributes about hosts from 3,rd,party tools via API,Attributes can be used in scheduling&placement policies,Useful to organize hosts by location,energy efficiency,model types,lease dates,Supports both static and dynamic attribute updates,资产管理与监控,Asset,DB,Resource,Updater,Monitoring DB,ISF,CMDB,API,12/17/2024,26,资源使用数据用来做计费和能力规划,Rate,Definition,Usage Data,Reserved Amount&Time,On-Demand Amount&Time,CPU Usage,Mem Usage,By Account,By Application Def,By User,By Resource Group,Platform ISF,HPC Application,Reporting App,JBOSS App,External Billing System,Other Data,12/17/2024,27,存储的集成,10G,10TB,Storage Containers,5G,100G,5TB,20TB,2G,10G,Storage,Adaptor,ISF Master,25G,50G,适配器来发现存储的内容和生成存储的资源池,用户使用存储资源就如同使用一个服务的定义,系统可以预约存储资源,系统启动适配器来生成存储的分配空间,工作流和后置配置步骤用来连接存储服务器,Storage Manager APIs(EMC,NetApp),High Performance,Mid-range storage,Low-end storage,12/17/2024,28,网络集成,ISF,Physical,Device,IP Pool Management,LB,Adapter,FW,Adapter,SW,Adapter,Network Device Adapter,IP Pool 1,IP Pool N,Physical,Device,Physical,Device,Physical,Device,Physical,Device,External Pool,Router,Adapter,Network Adapter,12/17/2024,29,企业私有云,West Coast,Data Center,East Coast,Data Center,Europe,Data Center,ISF Market Place Broker,Enterprise Messaging Infrastructure/JMS,Enterprise DB,Infrastructure,Remote Proxy,Leverage EGO capabilities to distribute ISF services across multiple nodes and provide centralized monitoring,troubleshooting,and HA,Application,End-User,Cloud,Admin,ISF Master Node,P-VMO,Portal,ICM,CES,App Manager,Blad,eLogic,VMware,Amazon,Tivoli,xCAT,ISF Master Node,P-VMO,Portal,ICM,CES,Blad,eLogic,VMware,Citrix,RedHat,Opsware,xCAT,VMware,xCAT,Nagios,Remote Proxy,VMware,xCAT,Nagios,12/17/2024,30,Platform Computing,Inc.,The leader in cluster,grid and cloud management software:,17 years of profitable growth,2,000 of the world,s most demanding client organizations,5,000,000 CPUs under management,500 professionals working across 13 global centers,Many Strategic relationships including Dell,Intel,Microsoft,Red Hat and SAS,Platform,Clusters,Grids,Clouds,Whatever,Computing,12/17/2024,31,Workload Management,Platform Computing,Resource Management,Clusters,Platform Cluster,Platform MPI,Grids,Platform LSF,Platform Symphony,Clouds,Platform ISF,Platform Product Portfolio,12/17/2024,32,Thank you,zfb,