Issues in Heterogeneous Network Based Concurrent Computing

基于异构网络的并发计算问题

• 批准号: 9118787
• 负责人: Vaidy Sunderam
• 金额: $ 13.93万
• 依托单位: Emory University
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 1992
• 资助国家: 美国
• 起止时间: 1992-01-15 至 1995-06-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9118787&HistoricalAwards=false
• 关键词: Issues; Heterogeneous; Network; Based; Concurrent

This project is concerned with concurrent computing on interconnected collections of computer systems, including multiprocessors, vector supercomputers, and specialized scalar machines. Such a network based, virtual concurrent computing environment is valuable both for increasing overall resource availability and sharing, and for exploiting special capabilities of individual machines. This project will investigate critical issues in such a computing system, and develop solutions that permit the effective and efficient utilization of network based concurrent processing environments. Analyses of high-performance applications, existing tools, and our experiences with a prototype network based system have identified the key areas that will be focussed upon in this research. One concerns appropriate methodologies and tools to assist in mapping portions of the application to the best suited resources in a heterogeneous environments. Another area of importance concerns improving the efficiency of communications facilities in networked environments. Support mechanisms to enable efficient concurrent computing over different network types, strategies for the implementation high level concurrency primitives, and lowering the overheads of operating in heterogeneous environments are issues that will be focused upon. The project will have a pragmatic orientation, and an experimental approach will be adopted.

该项目涉及互连计算机系统集合上的并发计算，包括多处理器、矢量超级计算机和专用标量机。 这种基于网络的虚拟并发计算环境对于提高整体资源可用性和共享以及利用各个机器的特殊功能都很有价值。 该项目将研究此类计算系统中的关键问题，并开发允许有效且高效地利用基于网络的并发处理环境的解决方案。 对高性能应用程序、现有工具以及我们在基于网络的系统原型方面的经验的分析已经确定了本研究将重点关注的关键领域。 其中之一涉及适当的方法和工具，以帮助将应用程序的各个部分映射到异构环境中最适合的资源。 另一个重要领域涉及提高网络环境中通信设施的效率。 在不同网络类型上实现高效并发计算的支持机制、实现高级并发原语的策略以及降低异构环境中的操作开销都是将重点关注的问题。 该项目将以务实为导向，采用实验性方法。