Unified Data Replication Architecture

国家科学基金会分布式系统研究CEISE/CNSABSTRACT论文编号：0411026主要研究员：Dahlin，Michael D.机构：德克萨斯大学奥斯汀分校Proposal标题：统一数据复制体系结构数据复制是提高大规模分布式系统性能和可用性的基本技术。然而，当前的复制机制与政策决策纠缠在一起，使得在新的环境中重新使用现有系统变得困难，并限制了使用这些机制构建的系统的有效性。这项研究的目标是为大规模分布式服务开发一个统一的数据复制架构和工具包。如果成功，这项工作将简化系统的构建，提高系统设计人员可用的可用性与一致性和性能之间的权衡，并让社区对数据复制核心技术有更深入的了解。具体的研究目标包括：(1)开发一种新的大规模分布式系统数据复制体系结构，将复制机制与策略分离；(2)系统地开发改进的底层复制机制、可扩展且灵活的一致性机制和自调优数据放置机制；(3)构建一个用于技术精化和迁移的工具包，并使用该工具包开发几个关键的案例复制系统。布雷特·D·弗莱施项目主任，CEISE/CNS，2004年5月28日。

National Science FoundationDistributed Systems Research CISE/CNSABSTRACTProposal Number: 0411026Principal Investigator: Dahlin, Michael D.Institution: University of Texas at AustinProposal Title: Unified Data Replication ArchitectureData replication is a fundamental technique for improving the performance and availabil-ity of large-scale distributed systems. However, current replication mechanisms are en-tangled with policy decisions that make it difficult to re-use existing systems in new envi-ronments and that restrict the effectiveness of systems built using these mechanisms. The goal of this proposed research is to develop a unified data replication architecture and toolkit for large-scale distributed services. If successful, this work will will simplify building systems, improve the availability v. consistency v. performance trade-offs avail-able to system designers, and give the community a deeper understanding of the tech-niques at the core of data replication. Specific research objectives include (1) developing a novel architecture for data replication in large scale distributed systems that separates replication mechanism from policy, (2) systematically developing improved low-level replication mechanisms, scalable and flexible consistency mechanisms, and self-tuning data placement mechanisms, and (3) constructing a toolkit for technology refinement and transfer as well as using this toolkit to develop several key case study replication systems.Dr. Brett D. FleischProgram Director, CISE/CNSMay 28, 2004.