CSR-PDOS-Collaborative: Virtual Machines Meet Application Clusters: A Highly Responsive Global Utility Computing Platform for Internet Applications

协同：虚拟机与应用集群的结合：互联网应用的高响应全球效用计算平台随着互联网应用的无所不在，面向服务的计算在全球计算中日益占据主导地位。因此，理解如何为面向服务的计算构建高效的平台对于支持互联网的持续发展至关重要。本研究的首要目标是获得这样的理解，特别是研究构建全球互联网应用效用计算（UCIA）平台的基本原则，该平台在资源重新分配方面具有高响应性，为不同的应用程序提供资源隔离，并支持异构执行环境。这项研究包括若干具体的方向。这包括研究在全球UCIA平台上资源共享的不同架构之间的权衡，以及研究有效的算法方法来请求分发和应用程序放置，这些方法考虑了执行算法的计算成本和这两个策略的交互。它特别关注全球UCIA的两种基础技术，广域应用集群和虚拟机之间的交互。通过以一种新颖的方式利用这些技术，本研究有望显著加快应用程序之间的资源再分配。为了量化其结果，该项目利用了一个广域测试平台，运行现成的Internet应用程序样本，并模拟了几个数据中心。Internet应用正在将Web从一个信息系统转变为一个面向服务的全球计算平台。通过提高我们为互联网应用构建高响应、全球效用计算平台的能力，这项研究将支持在这一重要方向上的持续进步。

CSR-PDOS-Collaborative: Virtual Machines Meet Application Clusters: AHighly Responsive Global Utility Computing Platform for InternetApplicationsWith the omni-presence of Internet applications, service-oriented computingincreasingly plays a dominant part in global computing. Thus,understanding how to build efficient platforms for service-orientedcomputing is critical in supporting continued Internet proliferation. Theoverriding goal of this research is to gain such understanding, and inparticular to investigate basic principles in building a global UtilityComputing for Internet Applications (UCIA) platform that is highlyresponsive in resource reallocation, provides resource isolation todifferent applications, and supports heterogeneous execution environments.This research encompasses a number of specific directions. This includesstudying tradeoffs between different architectures for resource sharing ina global UCIA platform and investigating efficient algorithmic approachesto request distribution and application placement that account for thecomputational cost of executing the algorithms and the interaction of thesetwo policies. It focuses in particular on the interaction between twofoundational technologies in global UCIA, wide-area application clusteringand virtual machines. By leveraging these technologies in a novel way,this research promises a significant speed-up in resource redistributionamong applications. To quantify its findings, this project utilizes awide-area testbed running a sample of off-the shelf Internet applicationsand emulating several data centers.Internet applications are transforming the Web from an information systeminto a global service-oriented computing platform. By improving ourability to build highly responsive, global utility computing platforms forInternet applications, this research will support continued progress inthis important direction.