Assignment and Allocation of Parallel Processing Systems

本项目将研究分布式内存并行处理器的作业调度技术。更具体地说，该项目将评估在多个独立应用程序之间共享多计算机系统的处理器和其他资源的几种策略的性能。评估将突出各种方法之间的权衡，并确定一种策略优于其他策略的条件。评估将侧重于分析方法和使用模拟的实验评估。在评估期间，通常在多台计算机上执行的各种应用程序将用作系统工作负载。该项目的结果将对未来消息传递MIMD系统的操作系统级调度有价值。＊＊＊

This project will study job scheduling techniques for distributed memory parallel processors. More specifically, the project will evaluate the performance of several strategies for sharing the processors and other resources of a multicomputer system among multiple, independent applications. The evaluation will highlight the tradeoffs among the approaches and identify the conditions under which a strategy outperforms the others. The evaluation will focus on both analytical methods and experimental evaluation using simulations. Various applications that are typically executed on multicomputers will be used as system workload during the evaluation. The results of the project will be valuable to OS level scheduling on future message passing MIMD systems. ***