Distributed, Self-Stabilizing Tasking for Emerging Network Environments

国家科学基金分布式系统研究CISE/CNSABSTRACTPROPOSAL NUMBER: 0411047主要研究者：Rubenstein， DanINSTITUTION：哥伦比亚大学提案题目：新兴网络环境的分布式自稳定任务本提案探讨分布式算法和协议在新兴网络环境中执行任务。任务分配很容易被非正式地描述为从相对较小的可能性集中抽取一个作业、资源或配置分配给每个网络节点，其中相同的任务可以分配给多个节点。任务示例包括802.11 ad-hoc网络的信道分配、GRID网络中的分布计算以及在P2P网络中放置内容副本。该项目包含一个重要的理论组成部分，它形式化了分布式算法和协议，证明了自稳定性，并形式化地量化了在一般网络拓扑上任务结果的性能界限。分布式协议是为实际实现而设计的，也就是说，设计的简单性优先于额外的最优性。我们计划通过在几个新兴网络环境（包括ad-hoc网络和分布式游戏网络）中对理论工作的结果进行模拟和原型设计来证明这一点。我们期望通过在主要会议和期刊上发表结果，以及软件原型和模拟包的公开发布来传播。该建议解决了几个短期目标，包括减少消息传递开销、底层节点加入和离开网络的协议性能，以及最优配置的变体，如内容流行度和任务硬连接的节点。该研究还调查了任务在两种新环境中的应用：在多跳自组织无线网络中分配信道，以尽量减少相邻节点之间的干扰，以及在分布式游戏基础设施中向游戏服务器分配任务，以尽量减少延迟。Brett D. fleisch CISE/ cns2004年5月26日

National Science FoundationDistributed Systems Research CISE/CNSABSTRACTPROPOSAL NUMBER: 0411047PRINCIPAL INVESTIGATOR: Rubenstein, DanINSTITUTION: Columbia UniversityPROPOSAL TITLE: Distributed, Self-Stabilizing Tasking for Emerging Network EnvironmentsThis proposal explores distributed algorithms and protocols to perform tasking in emerging networking environments. Tasking is easily described informally as assigning to each network node a job, rsource, or configuration drawn from a relatively small set ofpossibilities, where the same task may be assigned to multiple ndes. Tasking examples include channel assignment for 802.11 ad-hoc networks, distributing computations in GRID networks, and placing content replicas in P2P networks.This project contains a significant theoretical component that formalizes the distributed algorithms and protocols, proves self-stability, and formally quantifies bounds on the performance of the tasking result over general network topologies. The distributed protocols are designed for practical implementation, i.e., simplicity of design is preferred over additional optimality. We plan to demonstrate this facet via simulation and prototyping the results from the theoretical work within several emerging network environments including ad-hoc networks and distributed gaming networks. We anticipate dissemination via publication of results in leading conferences and journals, plus public release of software prototypes and simulation packages.Several short-term goals are tackled under this proposal, including the reduction of messaging overhead, performance of the protocols in which underlying nodes join and leave the network, and variants on optimal configuration such as content popularity and nodes whose tasks are hard-wired. The research is also investigating the application of tasking in two novel environments: channel allocation in multi-hop ad-hoc wireless networks to minimize interference among neighboring nodes, and assignment of tasks to gaming servers to minimize latency in distributed gaming infrastructures.Dr. Brett D. FleischProgram Director, CISE/CNSMay 26, 2004..