CSR--AES: Creating a Robust Desktop Grid using Peer-to-Peer Services

在最近的几篇论文中，对等计算和网格计算的融合被描述为不可避免的。P2P(Peer-to-Peer)服务显示出在许多领域提供健壮性、可靠性和分散性的前景，而网格计算已经显示出巨大的好处，使应用程序能够利用整个Internet上机器上可用的空闲周期。在这个项目中，我们使用P2P服务来允许用户(对等点)提交要在系统中运行的作业，以及运行其他用户提交的作业对系统中可用的任何资源进行操作。所有节点都向一个特别资源池贡献资源，并且所有节点都可以跨机构边界提交使用可用资源执行的作业。使用P2P服务可以提供一个健壮、可靠、可扩展的作业提交和执行系统，该系统能够利用一组用户愿意共享的任何计算资源。我们正在研究分布式算法，用于提交作业并有效地将它们与可用的资源进行匹配。我们使用P2P技术来实现负载平衡和弹性。从初步的分析和实验来看，我们的方案随着系统规模的扩大而扩展，并且对组件故障和对等离开/加入具有健壮性。这项工作是计算机科学家和天文学家共同努力的结果。我们正在使用一组来自计算天文学的应用程序和相关工作负载来设计、实现和验证算法。

The confluence of peer-to-peer and Grid computing has been described asinevitable in several recent papers. Peer-to-peer (P2P) services show thepromise of providing robustness, reliability, and decentralization in manyareas, while Grid computing has shown great benefits for enablingapplications to take advantage of idle cycles available on machines allacross the Internet. In this project we are employing P2P services to bothallow users (peers) to submit jobs to be run in the system and to run jobssubmitted by other users on any resources available in the system. Allpeers contribute resources to an ad-hoc resource pool, and all peers cansubmit jobs that are executed using available resources, acrossinstitutional boundaries. Using P2P services can provide a robust,reliable, scalable job submission and execution system that is able to takeadvantage of any computational resources that a set of users are willing toshare.We focus on the crucial problem of completely distributed and decentralizedjob placement in a Grid environment. We are working on distributedalgorithms for submitting jobs and efficiently matching them to availableresources. We use P2P techniques for both load balancing and forresilience. From preliminary analysis and experiments, our schemes bothscale with system size and are robust against component failures and peerdepartures/joins. The work is a collaborative effort between computerscientists and astronomers. We are designing, implementing, and validatingthe algorithms using a set of applications and associated workloads fromcomputational astronomy.