《Google云计算部署介绍》由会员分享,可在线阅读,更多相关《Google云计算部署介绍(67页珍藏版)》请在金锄头文库上搜索。
1、Cloud ComputingPing Yehg June 14, 2008Evolution of Computing with the NetworkNetworkNetwork ComputingNetwork is computer (client - server)Separation of FunctionalitiesCluster ComputingSeparation of FunctionalitiesGrid ComputingUtility ComputingypgCluster and grid images are from Fermilab and CERN, r
2、espectively.Evolution of Computing with the NetworkNetworkNetwork ComputingNetwork is computer (client - server)Tightly coupled computing resources:Separation of FunctionalitiesCluster ComputingTightly coupled computing resources: CPU, storage, data, etc Usually connected within a LAN Managed as a s
3、ingle resourceSeparation of FunctionalitiesManaged as a single resource Commodity, Open SourceGrid ComputingUtility ComputingypgCluster and grid images are from Fermilab and CERN, respectively.Evolution of Computing with the NetworkNetworkNetwork ComputingNetwork is computer (client - server)Tightly
4、 coupled computing resources:Separation of FunctionalitiesCluster ComputingTightly coupled computing resources: CPU, storage, data, etc Usually connected within a LAN Managed as a single resourceSeparation of FunctionalitiesManaged as a single resource Commodity, Open SourceGrid ComputingResource sh
5、aring across administrative domains Decentralized, openUtility ComputingDecentralized, open standards, non-trivial serviceGlobal Resource SharingypgCluster and grid images are from Fermilab and CERN, respectively.Evolution of Computing with the NetworkNetworkNetwork ComputingNetwork is computer (cli
6、ent - server)Tightly coupled computing resources:Separation of FunctionalitiesCluster ComputingTightly coupled computing resources: CPU, storage, data, etc Usually connected within a LAN Managed as a single resourceSeparation of FunctionalitiesManaged as a single resource Commodity, Open SourceGrid
7、ComputingResource sharing across administrative domains Decentralized openUtility ComputingDont buy computers lease computingDecentralized, open standards, non-trivial serviceGlobal Resource SharingypgDont buy computers, lease computing power Upload, run, downloadOwnership ModelpCluster and grid ima
8、ges are from Fermilab and CERN, respectively.The Next Step: Cloud ComputingServices and data are in the cloud,Services and data are in the cloud, accessible with any device connected to the cloud with a browserThe Next Step: Cloud ComputingServices and data are in the cloud,Services and data are in
9、the cloud, accessible with any device connected to the cloud with a browserA kth il ifdlA key technical issue for developers:ScalabilitScalabilityApplications on the WebYour Coolest Web ApplicationYour userinternet splat map: http:/ CC-by 2.0 baby picture: http:/ CC-by-sa 2.0internet splat map: http
10、:/ CC-by 2.0 baby picture: http:/ CC-by-sa 2.0Applications on the WebYour Coolest Web ApplicationThe CloudYour userinternet splat map: http:/ CC-by 2.0 baby picture: http:/ CC-by-sa 2.0internet splat map: http:/ CC-by 2.0 baby picture: http:/ CC-by-sa 2.0松下問童子松下問童子言師採藥去言師採藥去只在此山中只在此山中雲深不知處賈島尋隱者不遇I a
11、sked the kid under the pine treeI asked the kid under the pine tree, “Where might your master be?“ “He is picking herbs in the mountain,“ he said, “the cloud is too deep to know where “the cloud is too deep to know where. Jia Dao, “Didnt meet the master,“ written around 800ADHow many users do you wa
12、nt to have?Your Coolest Web ApplicationApplicationThe CloudThe CloudHow many users do you want to have?ThCldThe CloudYour Coolest Web ApplicationGoogle GrowthN98 10 000i25tNov. 98: 10,000 queries on 25 computers Apr. 99: 500,000 queries on 300 computers Sep. 99: 3,000,000 queries on 2100 computersSl
13、 bilitttScalability mattersCounting the numbersClient / ServerPersonal ComputerOne : ManyOne : OneCounting the numbersClient / ServerPersonal ComputerCloud ComputingOne : ManyOne : OneMany : ManyDeveloper transitionDeveloper transitionWhat Powers CloudWhat Powers Cloud Computing?Commodity Infrastruc
14、ture Softwarepg Hardware/foo/barPfilhitchunk . chunk . chunk . chunk . Performance: single machine not interesting Reliability:Distributed storage: Google File System (GFS) oMost reliable hardware will still fail: fault-tolerant software needed oFault-tolerant software enables useDistributed semi-st
15、ructured data system: BigTableof commodity components Standardization: use standardized machinessystem: BigTableDistributed data processing tMR dto run all kinds of applicationssystem: MapReducegoogle.stanford.edu (circa 1997) (1999) “cork bd “boards“Google Data Center (circa 2000) (new data center 2001) (3 days later) Current DesignIhk di In-house rack design PC-class motherboards Low-end storage andLow end storage