Improving Reliability and Availability of Chip Multiprocessors

提案ID：0811467提案标题：提高芯片多处理器的可靠性和可用性调查人员：马萨诸塞大学的桑迪普·昆杜，威斯康星大学的凯瓦尔·K·萨卢贾？威斯康星大学的麦迪逊·詹姆斯·E·史密斯？麦迪逊建议期：2008年6月至2010年5月总结：这是一个多所大学的项目。该方案的中心思想是开发和使用微视层。驻留在操作系统和微处理器的硬件之间。目标是即使在系统中的某些组件无法达到其指定目标时，也能提供所需的系统性能。特别是，本研究的重点是使用微振子层来保护处理器免受过热或用于测试和诊断处理器中的故障，以提高微处理器和基于微处理器的系统的可靠性和可用性。该项目将量化实施这一计划的优势和成本。这项研究的更广泛的技术和学术影响是通过网络共享将这项研究的结果提供给学术界和工业界。这一结果将对半导体行业大有裨益，因为该提案解决了迫在眉睫的可靠性问题。创新将以更低的成本生产出更高性能、高可靠性的计算资源，从而为社会提供巨大的社会经济效益，从而提高美国的竞争力。

Summary StatementProposal ID: 0811467 Proposal Title: Improving Reliability and Availability of Chip MultiprocessorsInvestigators: Sandip Kundu of University of Massachusetts Kewal K. Saluja of University of Wisconsin ? Madison James E. Smith of University of Wisconsin ? MadisonProposal Period: June 2008 to May 2010 Summary: This is a multi-university project. The central idea of the proposal is to develop and use a ?microvisor layer? that resides between operating system and the hardware of a microprocessor. The objective is to provide the desired system performance even when some of the components in the system are unable to meet their specified goals. In particular this study focuses on the use of microviosor layer to protect the processor from overheating or for testing and diagnosis of faults in the processor so as to improve the reliability and availability of a microprocessor and microprocessor based system. The project will quantify the advantages and the cost of implementing such a scheme.The broader technological and academic impact of this research is availability of results of this research to academia and industry though web share. The results will be of great benefit to the semiconductor industry as the proposal addresses reliability problems that are just around the corner. Innovations will help US competitiveness by producing higher performance, high reliability computing resources at lower cost, thus providing a great socioeconomic benefit to the society.