Improving the Robustness of Cluster-based Network Services Using Peer-to-peer Design Principles

使用对等设计原则提高基于集群的网络服务的鲁棒性

• 批准号: 0306473
• 负责人: Kai Shen
• 金额: $ 24.99万
• 依托单位: University of Rochester
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2003
• 资助国家: 美国
• 起止时间: 2003-08-01 至 2007-07-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0306473&HistoricalAwards=false
• 关键词: Improving; Robustness; Cluster; based; Network

Cluster-based network servers are a cost-effective platform forconstructing large-scale network services. However, service clustering presentssignificant challenges to system designers. Many components of the clusteringsupport are often implemented using centralized or hierarchical architectures, inorder to meet desired performance, consistency, and policy requirements. Theeffectivenessof these approaches has been demonstrated under well-controlledenvironments,but customized enhancements are often needed to deal with transientcomponent failures and fluctuating service loads.Peer-to-peer systems are distributed systems without any centralizedcontrolor hierarchical organization, and with functionally equivalent softwarerunningat each node. The primary goal of this project is to improve therobustness ofcluster-based network services using peer-to-peer system designprinciples.Specifically, this effort exploits three main principles of peer-to-peersystemdesign: functionally symmetric architecture, slow-scaling overhead, andnon-determinism. This project covers various components in serviceclusteringsupport, including service location lookup, membership management,cluster-wideresource management, and replication consistency support.Techniques investigated in this project can greatly benefit networkservicedesigners, for whom it can offer significant improvements in systemrobustness.In a broader context, this research complements the ongoing efforts ofbuildingglobal-scale peer-to-peer systems and services. In addition, thisprojectdirectly benefits systems-area instruction improvement at the PI'sinstitution.

基于集群的网络服务器是构建大规模网络服务的高性价比平台。然而，服务集群给系统设计人员带来了巨大的挑战。集群支持的许多组件通常使用集中式或分层架构实现，以满足所需的性能、一致性和策略要求。这些方法的有效性已经在良好控制的环境下得到了证明，但通常需要定制增强来处理瞬时组件故障和波动的服务负载。对等系统是分布式系统，没有任何集中控制或分层组织，并且在每个节点上运行功能等价的软件。该项目的主要目标是使用对等系统设计原则来提高基于集群的网络服务的健壮性。具体地说，这项工作利用了对等系统设计的三个主要原则：功能对称的体系结构、缓慢扩展的开销和非确定性。该项目涵盖了服务集群支持中的各个组件，包括服务位置查找、成员管理、集群范围内的资源管理和复制一致性支持。该项目中研究的技术可以极大地帮助网络服务设计人员，它可以显著提高系统的健壮性。在更广泛的背景下，本研究补充了正在进行的构建全球规模的对等系统和服务的努力。此外，该项目还直接惠及了PI机构的系统领域教学改进。