Allocation Shrinking Mechanisms for Improving Resource Utilization in Mixed Parallel Workflow Scheduling

Chih-Kai Lin ； Kuo-Chan Huang

《TANET2017 臺灣網際網路研討會》 2017年 (2017/10) Pp. 1096-1101

https://doi.org/10.6728/TANET.201710.0191

摘要

This paper explores the mixed-parallel workflow scheduling problem, which is becoming ever important because of the increasing large-scale, complex, and performance-demanding computational applications. Processor allocation is a crucial step in scheduling mixed-parallel workflows. We propose and evaluate a new allocation shrinking mechanism for improving resource utilization and thus reducing workflow execution makespan. Compared to a most recent previous approach, IAES, our new allocation shrinking mechanism can effectively prevent the inaccurate concurrency problem and features lower computational complexity. We have conducted a series of simulation experiments across two workflow types, various workflow sizes, and different numbers of processors. The experimental results show that our new allocation shrinking mechanism outperforms IAES significantly in almost all cases, especially for larger workflows.

關鍵字

mixed-parallel workflow scheduling ； processor allocation ； allocation shrinking ； resource utilization