arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2207.06902cs.DCcs.SC
收藏

YewPar中的并行Flowshop

Parallel Flowshop in YewPar

Ignas Knizikevičius, Phil Trinder, Blair Archibald, Jinghua Yan

摘要:

对于许多运筹学(OR)问题,并行组合搜索可以减少找到精确解的时间,但并行组合搜索是极具挑战性的。YewPar是一个新的组合搜索框架,旨在让领域专家受益于并行重用复杂的并行搜索模式。本文表明:(1)对一个典型的OR问题进行编码和并行化的工作量很小(2)YewPar库使得利用三种替代FSP并行化非常容易;(3)YewPar FSP实现是有效的,并且具有与已发布算法相当的顺序性能;以及(4)在Beowulf集群上的10个标准FSP实例上提供了三个并行FSP版本的系统性能评估,该实例具有多达240个工作者。

英文摘要:

Parallelism may reduce the time to find exact solutions for many Operations Research (OR) problems, but parallelising combinatorial search is extremely challenging. YewPar is a new combinatorial search framework designed to allow domain specialists to benefit from parallelism by reusing sophisticated parallel search patterns. This paper shows (1) that it is low effort to encode and parallelise a typical OR problem (Flowshop Scheduling FSP) in YewPar even for scalable clusters; (2) that the YewPar library makes it extremely easy to exploit three alternate FSP parallelisations; (3) that the YewPar FSP implementations are valid, and have sequential performance comparable with a published algorithm; and (4) provides a systematic performance evaluation of the three parallel FSP versions on 10 standard FSP instances with up to 240 workers on a Beowulf cluster.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑