摘要
目前大规模的并行分布多处理机系统中,调度算法好坏直接影响计算系统的高性能计算潜力能否发挥,调度的目的就是如何分配资源使系统性能最优。本文主要讨论分布式多处理机系统进行任务调度时的关键问题,包括问题模型的描述,调度策略,常用算法,评估标准,数据平台以及该问题的发展趋势。
The scheduling algorithm in massively parallel system is a key factor to the potential power of high-performance systems. The purpose of scheduling is to control the allocation of resources and ensure optimal system performance. This paper analyses the key issues in distributed multiprocessor, including the model description, scheduling strategy, commonly used algorithm, evaluation criteria, data platform and future trend.
出处
《河南科技学院学报》
2007年第2期46-49,共4页
Journal of Henan Institute of Science and Technology(Natural Science Edition)
基金
河南省教育厅自然科学研究项目(200510467004)
关键词
多处理机
并行程序
有向无环图
任务调度
multiprocessor
parallel program
directed acyclic graph
task scheduling