【正文】 為了管理這種情況下，刀片式服務(wù)器內(nèi)積云須提供數(shù)百一些預(yù)安裝的軟件相同的虛擬機(jī)。3． 瀏覽可用的虛擬機(jī)映像。此外，后端服務(wù)器，可以實(shí)現(xiàn)一些復(fù)雜的。filetype=.taramp。雇用更專業(yè)的方式OpenNEbula，我們兼并了一些現(xiàn)代的如LDAP [13]和Oracle集群文件系統(tǒng)[30]安全基礎(chǔ)設(shè)施解決方案OpenNEbula。我們重新設(shè)計(jì)了Globus的虛擬工作區(qū)的服務(wù)，以適應(yīng)作為積云前端服務(wù)中。在主要的Globus虛擬工作空間服務(wù)可受聘為云服務(wù)前端。我們用一個(gè)虛擬機(jī)模板管理工具操作系統(tǒng)農(nóng)場(chǎng)。3 積云：以科學(xué)的云積云項(xiàng)目是一項(xiàng)持續(xù)的云在最近成立的Steinbuch中心計(jì)算（鱗癌）在德國卡爾斯魯厄理工學(xué)院（KIT）的計(jì)算項(xiàng)目。虛擬數(shù)據(jù)系統(tǒng)（VDS）的[35]可能是很好的參考。?世界級(jí)分布式存儲(chǔ)系統(tǒng) 云存儲(chǔ)模型應(yīng)當(dāng)預(yù)見：網(wǎng)絡(luò)存儲(chǔ)系統(tǒng)，它是由分布式存儲(chǔ)提供商（如數(shù)據(jù)中心），為用戶提供支持租用的存儲(chǔ)容量。服務(wù)機(jī)構(gòu)和業(yè)務(wù)流程可能是云層內(nèi)部管理在面向服務(wù)架構(gòu)（SOA）。?提供QoS保證通過計(jì)算云計(jì)算環(huán)境可以提供保證用戶的QoS，例如，硬件如CPU速度，I / O帶寬和內(nèi)存大小的表現(xiàn)。 主要特點(diǎn)因此，減輕了SaaS的客戶對(duì)軟件維護(hù)的負(fù)擔(dān)，降低了軟件采購費(fèi)用由需求定價(jià)。根據(jù)我們的經(jīng)驗(yàn)，我們提出了一個(gè)早期的云計(jì)算的定義如下 ：一個(gè)計(jì)算云是啟用服務(wù)的網(wǎng)絡(luò)設(shè)置，提供可擴(kuò)展，服務(wù)質(zhì)量保證，通常個(gè)性上的需求，這可以在一個(gè)簡單和普遍的方式訪問廉價(jià)的計(jì)算平臺(tái)。類似的情況發(fā)生在50年前：分時(shí)計(jì)算服務(wù)器為多個(gè)用戶服務(wù)。在本文中，我們目前的積云計(jì)算項(xiàng)目。group=baseWe have built an OS Farm client and embedded it into the Cumulus frontend. When users require virtual machine images, the Cumulus frontend service invokes a wget mand to generate virtual machine images for users. This implementation frees users from the pain to generate and submit virtual machine images manually. Networking solutionWe provide a new networking solution: the “forward”mode. Users do not have to specify anything for the network requirements. The OpenNEbula starts the Xen virtual machine images, allocates dynamic IP addresses for the virtual machines and then returns them to the users. In addition the backend servers could implement some plex ? Support a new networking solution – the forward network management policies and return IP addresses to the users. The new networking solution is transparent to users, thus it is named as the “forward” mode. Access to the Cumulus serviceThe Cumulus service provides the following access? via Globus virtual workspace service client or Nimbus cloudkit client,? with Grid puting workbench or existing Grid portals, like gEclipse [10], Open Grid Computing Environment [15] and GridShell [16].The gEclipse project [9] is an international research effort, which aims to build an integrated workbench framework to leverage the power of existing Grid infrastructures. The gEclipse project plans to extend itself to access our Cumulus Cloud service by the following steps:1. create a VO for the Cumulus user, 2. provide a Grid credential,3. browse the available virtual machine images,4. launch the desired virtual machines,5. access the virtual machine.We also expect to integrate the OS Farm into the gEclipse platform using its generic Grid connection concept. In the gEclipse project, a connection allows the user to link to a local or remote ?le system. To establish a connection, users need only to provide the necessary. Typical use casesIn the following we present several typical use cases and patterns for Cumulus usage:? GridKa school training supportResearch Center Karlsruhe yearly holds the GridKa school, which is a training session covering topics related to Scientific Grid Computing. In GridKa school, more than 100 participants take courses and practice various Grid middle wares. To manage such a scenario, the server blades inside Cumulus are required to provide several hundreds identical virtual machines with some preinstalled software. Each virtual machine is allocated a dynamic IP address and the trainees are able to access their own virtual machine directly.? Virtual eScience infrastructureUsers could dynamically organize their customized eScience infrastructure on the distributed virtual machines provided by Cumulus. The OS Farm server provides prototype virtual machines which are then enriched with the requested software configuration, for example, an eScience infrastructure could contain GridSAM [24] as Grid middleware, ActiveBPEL engine [28] as work?ow engine to build a work?ow system. With these preconfigured virtual machine images, users could build an eScience infrastructure at runtime by multiple service orchestration and invocation.? Virtual puting centerUsers could lease a large set of virtual machines from the Cumulus, deploy a middleware on the distributed virtual machines to build a virtual puting center, for example, a plete UNICORE [19] installation. A virtual puting center functions in the same way as a real puting center. It provides portals or Web services via which users can access puting resources in the virtual puting center. The builder of the virtual puting center, who normally works as the administrator of the virtual center, provides application level software and libraries on the puting resources and defines the management policies and access control rules.4 ConclusionThis paper reviews the recent advances of Cloud puting and presents our early results from the Cumulus Cloud puting project. The Cumulus project aims to build ascalable scientific cloud: Cumulus could adaptive to various virtual machine hypervisors as backend, furthermore delegates users’requirements to other Clouds. This objective defines the resource reorganization and service orchestration for Cloud puting.References [1] P. Barham, B. Dragovic, K. Fraser, S. Hand, T. L. Harris, A. Ho, R. Neugebauer, I. Pratt, and A. and the art of virtualization. In Proceedings of the 19th ACM Symposium on Operating Systems Principles, pages 164–177, New York, U. S. A., Oct. 2003.[2] Here es HaaS [URL]., access on June 2008.[3] J. Dean. Mapreduce and other building blocks for largescale distributed systems at google. In Proceedings of the USENIX Annual Technical Conference,2007.[4] J. Dean and S. Ghemawat. Mapreduce: simplified data processing on large clusters. Commun. ACM,51(1):107–113, 2008.[5] Web definition[URL]., access on June 2008.[6] B. Gleeson etc. A framework for ip based virtual private networks. Rfc2764, The Internet Engineering Task Force, Feb. 2000.[7] G. Fedak, C. Germain, V. Neri, and F. : A generic global puting system. In Proceedings of the 1st IEEE International Symposium on Cluster Computing and the Grid, pages 582–