SEParAT: scheduling support environment for parallel application task graphs

SEParAT: scheduling support environment for parallel application task graphs
复制标题

DOI:
10.1007/s10586-012-0211-1
复制
发表时间:
2012-09
期刊:
Cluster Computing
影响因子:
--
通讯作者:
Jörg Dümmler;Raphael Kunis;G. Rünger
Jörg Dümmler;Raphael Kunis;G. Rünger
中科院分区:
其他
文献类型:
--
作者:
Jörg Dümmler;Raphael Kunis;G. Rünger

文献摘要

被引文献

相似文献

使用并行任务的程序可以由任务图表示,使得调度算法可以用于找到并行任务的有效执行顺序。本文提出了一个灵活的,基于组件的和可扩展的调度框架称为SEParAT,支持调度的并行程序在多种方式。本文介绍了SEParAT的功能和软件体系结构。灵活的接口支持与其他编程工具的协作,例如,开发应用程序的并行任务结构规范的工具。SEParAT的核心组件是一个可扩展的调度算法库,它提供了一个基础设施来确定任务图的有效调度。可以处理同构和异构平台。文章还包括详细的实验结果,包括SEParAT的评估以及各种调度算法的评估。
Programs using parallel tasks can be represented by task graphs so that scheduling algorithms can be used to find an efficient execution order of the parallel tasks. This article proposes a flexible, component-based and extensible scheduling framework called SEParAT that supports the scheduling of a parallel program in multiple ways. The article describes the functionality and the software architecture of SEParAT. The flexible interfaces enable the cooperation with other programming tools, e.g., tools exploiting a specification of the parallel task structure of an application. The core component of SEParAT is an extensible scheduling algorithm library that provides an infrastructure to determine efficient schedules for task graphs. Homogeneous as well as heterogeneous platforms can be handled. The article also includes detailed experimental results comprising the evaluation of SEParAT as well as the evaluation of a variety of scheduling algorithms.