1、Data GridsvDataGrid:A dynamic logical namespace that enables coordinated sharing of heterogeneous distributed storage resources and digital entities based on local and global policies across administrative domains in a virtual enterprise.vDataGridLogical name space for location independent identifie
2、rsAbstractions for storage repositories,information repositories,and access APIsLatency managementData Characteristics in GridvMostly unstructured data,heterogeneous resourcesImages,files,semi-structured,databases,streams,File systems,SAN,FTP sites,web servers,archivesvCommunity-BasedShared amongst
3、one or more communitiesvMeta-dataDifferent meta-data schemas for the same dataDifferent notations,ontologiesvSensitive to SharingData Grid TransparenciesvAccess data without knowing the type of storageStorage repository abstraction vFind data without knowing the identifierDescriptive attributesvAcce
4、ss data without knowing the locationLogical name spacevRetrieve data using your preferred APIAccess abstractionvProvide transformations for any data collectionData behavior abstractionLogical Layers(bits,data,information,.)Storage Resource TransparencyStorage Location TransparencyData Identifier Tra
5、nsparencyData Replica TransparencyVirtual Data TransparencySemantic data Organization(with behavior)Inter-organizational Information Storage ManagementStorage Resource TransparencyvStandard operations at storage repositories POSIX like operations on all resourcesvStorage specific operationsDatabases
6、-bulk metadata accessObject ring buffers-object based accessHierarchical resource managers-status and staging requestsDAIS-Requirements&Functionalities Key Access&Integration Functionalities:vPublication and DiscoveryvStatements(Data Operations)vStructured Data TransportvData TransformationvTransact
7、ionsvMetadatavManagement:Operation&PerformancevData ReplicationvConnections and SessionsvIntegrationDAIS Functional ScopingPublishing and DiscoveryvService discovery via a registryvData service registry structure and contentvData service descriptionvDatabase contents descriptionLogical and Physical
8、schema,physical characteristicsvDatabase capabilityLanguages,features,characteristicsvTerminology and structure of descriptions DAIS Functional ScopingStatementsvAccess to RDBMS,XML,other databasesvOperationsPrepared and dynamic statementsNative DML,DDL,Context,Procedures/PackagesScripted operations
9、vSynchronous&asynchronous statement interfacesPreparation/validation,application,deliveryvNotificationEvent and informational vNatural Query languageDAIS Functional ScopingStructured Data TransportvDelivery from one source to one or more specified destinations along a series of channelsvTemporary st
10、orage during transportvAlternative delivery modesStreaming,multicastvDelivery monitoring and notifications vSupport for different protocols for delivery along different channelsvSystematic methods of encryption or compression on selected channelsDAIS Functional Scoping Data TransformationvBefore,dur
11、ing,after statement executionvRestructuringFormatting,sequencingvSchema changeRestructuring,naming,constraintsvConversionUnits,coordinate system,algorithmic(e.g.Fourier)Composition with computational servicesDAIS Functional Scoping TransactionsvDBMS capability E.g.Set transaction,commit,rollbackEven
12、t and informational notificationvDistributed transactionsHeterogeneous two-phase commit vRelaxed transaction modelsCore activity service model,collaborationsvDefinition of new transaction modelsDAIS Functional Scoping Authentication,Access Control,AccountingvBasic cost modelDBMS capability,e.g.resou
13、rces usedvDelegation of fine grained access rights vAccess based role model DAIS Functional Scoping MetadatavTechnical metadataLocation,physical schema,data characteristics,owner,version,access methodsDatabase capabilities and extensionsvContextual metadataLogical schema,classifications,terminologie
14、s,ontologies,derived datavContext for schema mappingSchema conversion and evolutionDAIS Functional Scoping Management:Operations and PerformancevExploit DBMS capabilitiesRAMPSReorganisation,backup/recovery,user managementDAIS Functional Scoping Data ReplicationvExploit DBMS capabilitiesReplication a
15、nd synchronisation featuresvCore replication capability Data definition,manipulation,transport operationsDAIS Functional Scoping Connections and SessionsvFurther investigation requiredDAIS Functional Scoping IntegrationvHighly dynamic federation vAlternate source selection form available replicasvFa
16、cilitate optimisationvSemantic based integrationTraditional Distributed Database Management System and Their Limitationsv Challenges to Distributed Database Management System Scale Heterogeneity Distribution AutonomyTransparencyFederated DBsvData DistributionvNo Common SchemaA federated DBMS serves
17、as a middlewareSolved Some Problem partially:vHeterogeneity transparencyvDistribution transparencyTraditional Distributed Data Management System and Their Limitations Inter-Operations Using ODBC/JDBC/OLE DB传统的分布环境下数据管理系统和 网格环境中数据管理系统的区别特征特征传统分布环境下传统分布环境下 数据管理系统数据管理系统网格环境下网格环境下 数据管理系统数据管理系统开放性开放性需求和技
18、术：需求和技术：有一定确定性和封闭性有一定确定性和封闭性开放系统、开放技术开放系统、开放技术通用性通用性专门领域、专门技术专门领域、专门技术通用技术通用技术集中性集中性统一规划、集中控制统一规划、集中控制自然进化、非集中控制自然进化、非集中控制使用模式使用模式终端终端 或或 C/S服务模式服务模式标准化标准化领域标准或行业标准领域标准或行业标准通用标准（通用标准（+行业标准）行业标准）平台性平台性应用解决方案应用解决方案平台或基础设施平台或基础设施Grid-enable DatabasevDatabase Requirements of Grid ApplicationvThe most im
19、portant requirementsDBS must support the,Grid standards:relevant,existing and emerging,for example the Grid Security InfrastructurevOther important requirementsRetrieval(Grid IR)ScalabilityHandling unpredictable usageMeta-driven accessMultiple Database Federation Grid-enable DatabasevCollective view
20、 of Inter-organizational data Operations on grid spacevLocal autonomy and global state consistencyvCollaborative communitiesMultiple administrative domains or“Grid Zones”vSelf-describing and self-manipulating data Horizontal and vertical behaviorLoose coupling between data and behavior(dynamically)R
21、elationships between a digital entity and its Physical locations,Logical names,Meta-data,Access control,Behavior,“Grid Zones”.Need for Standard DGLDatabaseSQL121.EventHits.sqlUniversity of Gators121.EventThit.xmlNational LabGrid-Enable DatabaseXML based,Invoke OperationsSubset XQueryDGLDDL,DML,DQL信息
22、技术的演变Mainframe大型主机客户服务器Internet计算结构部门级服务器“目前目前 IT 部门运行效率十分低下，部门运行效率十分低下，通常只利用了总容量的小部分通常只利用了总容量的小部分”Frank E.GillettForrester 研究公司，2002年10月当前IT所面临的问题v信息孤岛为最大负载而配置伸缩性有限可用性 99.x%安全控制分散v成本不断上涨v影响所有行业降低IT 成本v昂贵的硬件设备v附加成本亦十分昂贵v单点故障v提供企业级服务成本高昂v低成本模块化设备v附加成本低v无单点故障v提供企业级服务成本低大型专用服务器Oracle 网格技术企业级网格计算意味着vCEO
23、s降低费用随用随付vIT 经理提高可用性提高服务质量vIT 管理人员提高自动化程度和生产率减少错误对开发人员和独立软件商意味着v无须更改程序代码v应用系统的管理功能更强处理过程自动化数据库自动化管理v利用共有服务功能如：身份认证等v硬件系统的成本更低Oracle 网格计算结构v计算机资源共享池v虚拟与信息提供v负载均衡v高质量服务v自动化网格控制网格控制存储网格存储网格数据库网格数据库网格应用服务器网格应用服务器网格基于规则的负载均衡技术v实现于数据库和应用服务器集群中v基于规则动态分配服务器资源v自动分派服务请求v一个服务器出现故障，处理过程自动分派到其他服务器上v处理量变化时，重新分配服务
24、器容量网格管理v统一管理和监控v利用标准的规则进行管理 系统配置 性能调整 安全控制v自动化处理应用系统多个系统现有应用系统不需要修改就可以利用网格技术 套装应用系统 客户定制的应用系统 所有的应用系统数据库集群和应用服务器集群：经受考验的成熟技术v全球成千上万用户v运行于所有平台上 Transaction Processing Council(TPC),www.tpc.org.As of December 8,2003:Sixteen-node HP Integrity rx5670 server cluster,each with 4 Itanium 2 1.5 GHz processor
25、s,1,184,893.38 tpmC,$5.52/tpmC,available April 30,2004.HP Integrity Superdome server with 64 Itanium 2 1.5 GHz processors,HP-UX 11iv2,1,008,144.49 tpmC,$8.33/tpmC,available December 31,20031,184,893数据库性能1,108,144As of November 4,2003:Oracle Database 10g Enterprise Edition,HP Integrity Superdome,1,00
26、8,144.49 tpmC,$8.33/tpmC,available 12/31/03.Source:Transaction Processing Council(TPC),www.tpc.org创TPC-C世界记录541,764423,414As of September 5,2003:Oracle Database 10g Enterprise Edition,HP Integrity Superdome,824,164.53 tpmC,$8.28/tpmC,available 12/31/03.Oracle Database 10g Enterprise Edition,HP 9000
27、Superdome Enterprise Server,541,673 tpmC,$10.69/tpmC,available 12/31/03.Oracle9i Database Enterprise Edition,HP 9000 Superdome Enterprise Server,423,414 tpmC,$15.64/tpmC,available 12/31/03Source:Transaction Processing Council(TPC),www.tpc.org比Oracle9i快28%As of September 5,2003:Oracle Database 10g En
28、terprise Edition,HP Integrity Superdome,824,164.53 tpmC,$8.28/tpmC,available 12/31/03.NEC Express5800/1320Xd,Oracle Database 10g Enterprise Edition,521,440.53 tpmC,$11.77/tpmC,available February 15,2004.HP Integrity rx5670,(Oracle Database 10g Standard Edition),136,110.98 tpmC,$4.09/tpmC,available D
29、ecember 31,2003.Source:Transaction Processing Council(TPC),www.tpc.org136,111 以及 Linux世界记录1,165(2个个cpu)应用服务器性能2003年9月12日:HP ProLiant ML370G3(2个CPU);Red Hat Linux AS 2.1;Oracle 应用服务器10g;431.26 TOPS双节点，$160.62/TOPS单节点。HP rx5670(4 CPUs);HP-UX 11i;BEA WebLogic Server 7.0;408.02 TOPS双节点,$1075.17/TOPS双节点。
30、来源:SPEC(http:/www.spec.org)?IBM创Linux SpecjAS2002世界记录1,037BEA(4个个cpu)Oracle公司的数据中心vOracle大学网格v外包业务网格网格控制网格控制存储网格存储网格数据库网格数据库网格应用服务器网格应用服务器网格如何开展网格之旅 自动化及优化自动化及优化合并合并标准化标准化经过实施标准化、自动化及合并，拥经过实施标准化、自动化及合并，拥有主机、有主机、Unix 和和 Windows 典型配置的典型配置的企业即能节省企业即能节省 8.5%到到 10.5%的构建数的构建数据中心的预算。据中心的预算。“”Gartner Research:Gartner Research:实时基础实时基础架构对架构对 IT IT 运营预算的影响运营预算的影响 (Donna Scott,John A.Donna Scott,John A.Oborn,Barbara Gomolski)Oborn,Barbara Gomolski)2003 2003 年年 7 7 月月 17 17 日日
