Dual-Layer Cooperative Error Control for Reliable Nanoscale On-Chip Interconnection Networks

用于可靠的纳米级片上互连网络的双层协作错误控制

• 批准号: 0925993
• 负责人: Paul Ampadu
• 金额: $ 29.01万
• 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: 2009
• 资助国家: 美国
• 起止时间: 2009-08-01 至 2015-07-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0925993&HistoricalAwards=false
• 关键词: Dual; Layer; Cooperative; Error; Control

The objective of this research is to investigate energy-efficient cooperative dual-layer (data link and network layers) error control methods for nanoscale networks-on-chip. The approach is to combine hop-to-hop error detection at the data link layer with adaptive routing and end-to-end forward error correction at the network layer.With respect to intellectual merit, this research addresses the problem that, with aggressive technology scaling, interconnects become increasingly susceptible to noise and they emerge as the critical bottleneck in meeting reliability and energy constraints in nanoscale systems-on-chip. Research in reliable interconnects has typically investigated solutions at a single layer, e.g. physical or electrical. This research considers the hardware efficiency of switch-level error detection combined with the reliability performance of network-layer forward error correction, with the goal of simultaneously improving reliability and energy use. In particular, this research seeks to: (i) create new efficient codes for error detection and correction across layers in a network-on-chip; (ii) devise innovative methods, using configurable spare wires and adaptive routing, to manage intermittent and permanent errors; and (iii) examine and develop methods to mitigate the impact of faults in the encoder/decoder circuits themselves.With respect to broader impact, this research has the potential to create a new reliable interconnect design approach that can accelerate advances in scalable multi-core systems-on-chip. This breakthrough could have significant societal and economic impact with the realization of reliable energy-efficient integrated systems for numerous applications, ranging from infrastructure monitoring to health care. Knowledge generated is disseminated through workshops, lectures and a public website. This research actively engages underrepresented minority and female students and extends outreach through international collaborations to broaden the global perspectives of U.S. students.

本研究的目的是探讨能量有效的合作双层（数据链路层和网络层）的错误控制方法的纳米级片上网络。 该方法是结合联合收割机跳到跳的错误检测在数据链路层与自适应路由和端到端的前向纠错在网络layer.With方面的智力merit，本研究解决的问题，积极的技术规模，互连变得越来越容易受到噪声和他们出现的关键瓶颈，在满足可靠性和能源的限制，在纳米级系统芯片。 可靠互连的研究通常在单层（例如，物理或电气）上研究解决方案。 该研究考虑了交换机级错误检测的硬件效率与网络层前向纠错的可靠性性能相结合，目标是同时提高可靠性和能量使用。 特别是，这项研究旨在：（i）创建新的高效代码的错误检测和纠正跨层的片上网络;（ii）设计创新的方法，使用可配置的备用线和自适应路由，以管理间歇性和永久性错误;及（iii）研究及发展方法，以减轻编码器/解码器电路本身故障的影响。至于更广泛的影响，该研究具有创建新的可靠互连设计方法的潜力，该方法可以加速可扩展的多核片上系统的进步。 这一突破可能会产生重大的社会和经济影响，实现可靠的节能集成系统，用于从基础设施监测到医疗保健等众多应用。 所产生的知识通过讲习班、讲座和公共网站传播。 这项研究积极参与代表性不足的少数民族和女学生，并通过国际合作扩大外展，以拓宽美国学生的全球视野。