High-Performance Stable Packet Switches

高性能稳定分组交换机

• 批准号: 0435303
• 负责人: Shivendra Panwar
• 金额: --
• 依托单位: Polytechnic University of New York
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2005
• 资助国家: 美国
• 起止时间: 2005-01-15 至 2008-12-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0435303&HistoricalAwards=false
• 关键词: Performance; Stable; Packet; Switches

National Science FoundationNETS - Research in Network Technologies and Systems CISE/CNSABSTRACTProposal Number: 0435303Principal Investigator: Panwar, Shivendra S.Institution: Polytechnic University of New YorkProposal Title: High-Performance Stable Packet SwitchesWhile switching capacity follows Moore's Law and doubles every 18 months, op-tical fiber line bandwidths double every 12 months. Consequently, packet switching technology continues to be one of the bottlenecks in the development of broadband networks. In this project, the research team has devised a new class of low complexity matching algorithms, i.e. Exhaustive Service Match with Hamiltonian Walk, which achieves stability and low packet delay, as opposed to cell delay, which was the focus of most previous research. Specifically, the fol-lowing two important issues are studied: (i) determination of the optimal packet delay for a matching algorithm, and (ii) options for the packet delay to increase sub-linearly with switch size under a general traffic pattern. This project will also address the class of switches that resolve the output contention by using the load balancing, where no scheduler is used, and re-sequencing is needed at the output. The total packet delay being investigated consists of the queuing delay and the re-sequencing delay. Another goal is to re-examine switch archi-tectures and their performance under more realistic long range dependent (LRD) traffic, and suggest improved architectures tuned to LRD traffic based on the insights thus derived.Dr. Admela JukanProgram Director, CISE/CNSAug 5, 2004.

美国国家科学基金会-网络技术与系统研究CISE/ cnsabstract提案编号：0435305首席研究员：Panwar， Shivendra s.机构：纽约理工大学提案标题：高性能稳定分组交换机交换机容量遵循摩尔定律，每18个月翻一番，光纤线路带宽每12个月翻一番。因此，分组交换技术仍然是宽带网络发展的瓶颈之一。在本项目中，研究团队设计了一类新的低复杂度匹配算法，即Hamiltonian Walk的穷穷服务匹配算法，它实现了稳定性和低分组延迟，而不是之前大多数研究的重点。具体而言，研究了以下两个重要问题：(i)确定匹配算法的最佳数据包延迟，以及（ii）在一般流量模式下，数据包延迟随交换机大小亚线性增加的选项。该项目还将解决通过使用负载平衡解决输出争用的交换机类，其中不使用调度器，并且需要在输出处重新排序。所研究的总包延迟包括排队延迟和重排序延迟。另一个目标是重新检查交换机架构及其在更现实的远程依赖（LRD）流量下的性能，并根据由此得出的见解提出针对LRD流量的改进架构。Admela jukan2004年5月5日CISE/ cns5项目主任。