Performance and Yield of Defect and Fault Tolerant Multi- processors

缺陷和容错多处理器的性能和良率

• 批准号: 8805586
• 负责人: Israel Koren
• 金额: $ 27.26万
• 依托单位: University of Massachusetts Amherst
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 1988
• 资助国家: 美国
• 起止时间: 1988-09-01 至 1992-02-29
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=8805586&HistoricalAwards=false
• 关键词: Performance; Yield; Defect; Fault; Tolerant

Issues on performance and yield in defect- and fault-tolerant multiprocessors are being addressed in this research. Two different classes of multiprocessors are being considered in the research: shared-memory multiprocessors and multicomputer systems with local memory. Issues relating to metrics for capturing the most important properties of multiprocessors and algorithms for calculating the yield and performance of multiprocessors are being studied. The benefits and drawbacks of increasing the level of integration in the design of multiprocessors are being investigated. This research will result in an ability to determine the best level of integration for a given defect- and fault-tolerant multiprocessors system. Improvements in technology have reduced the defect density as well as the failure rates of all components in multiprocessing systems. However, the large increase in the number of components per system has more than offset the increase in yield and reliability of a single component. It is desirable that a given system have graceful degradation in performance in the presence of faulty components. The proposed research addresses the important issues of estimating performance and yield of multiprocessors and the benefits and drawbacks of increased level of integration. Results obtained will allow the design of more reliable and fault-tolerant multiprocessors. Support is strongly recommended.

这项研究正在解决缺陷和容错多处理器的性能和产量问题。 研究中考虑了两种不同类别的多处理器：共享内存多处理器和具有本地内存的多计算机系统。 正在研究与捕获多处理器最重要属性的指标以及计算多处理器产量和性能的算法相关的问题。 人们正在研究提高多处理器设计集成度的优点和缺点。 这项研究将能够确定给定缺陷和容错多处理器系统的最佳集成水平。 技术的进步降低了多处理系统中所有组件的缺陷密度和故障率。 然而，每个系统组件数量的大幅增加远远抵消了单个组件产量和可靠性的增加。 希望给定系统在存在故障组件的情况下性能能够适度下降。 拟议的研究解决了估计多处理器性能和产量的重要问题以及提高集成度的优点和缺点。 获得的结果将允许设计更可靠和容错的多处理器。 强烈建议支持。