缩短程序的执行时间是并行处理的首要目标，有效的任务分配算法是实现这一目标的关键，对群机系统来说更是如此．文中针对并行语言中常用的并行范式——单并发任务簇提出了近优分配算法 Ｏ Ｐ Ｔ Ａ，并在群机系统上做了与 Ｍ Ｈ 算法的比照实验，结果表明较 Ｍ Ｈ 算法缩短程序执行时间 １０％ 左右 Shorten the execution time of the program is the primary target of parallel processing, the effective task allocation algorithm is the key to achieve this goal, especially for the cluster system. In this paper, we propose a near-optimal allocation algorithm, O P T A, for the parallel paradigm commonly used in parallel languages, and compare it with the M H algorithm in a cluster system. The results show that the M H algorithm shortens the procedure The execution time is about 10%