一种基于动态规划的并行构件资源选择算法

摘要

当前,多种硬件体系结构并存,传统的并行构件资源管理却不能充分利用并行构件的属性以适应底层硬件的多样性,可扩展性比较差.在研究基于CCA规范的并行构件执行集成平台的基础上,本文提出了一种层次化并行构件资源管理模型,并提出了一种基于动态规划的并行构件资源选择算法.实验结果表明,相对于通常使用的CPU频率优先、CPU Cache值优先与通信优先策略,本文提出的基于动态规划的并行构件资源选择算法不仅能够更加精确地选择计算节点资源,而且具有良好的可扩展性.%At present, various hardware architectures coexist, traditional parallel component resource management can not fully take advantage of parallel component properties to meet the diversity of the underlying hardware. At the same time the scalability of traditional parallel component is rather poor. In this paper, the research based on the specifications of CCA proposes a hierarchical model for parallel component resource management, then this paper mainly proposes the resource selection algorithm based on dynamic programming. Experimental results show that related to the tactics of CPU frequency priority, CPU Cache value priority and communication priority, the resource selection algorithm based on dynamic programming can select the resource of computing node more precise, and has a good scalability.