Scheduling Independent Tasks on Metacomputing Systems

Metacomputing is a convenient and powerful abstraction for dealing with the complexities that arise when managing and using a large collection of heterogeneous computational resources. One of the most fundamental characteristics of a metacomputing system is the algorithm it uses for the scheduling placement of jobs on processing nodes. We describe five schedule placement algorithms, and report on their success and failure modes when used to schedule job distributions. We investigate five different distributions of job execution time and the effects of predictability on the algorithms' performance. Our objective in this work is to develop a hierarchical scheduling model for large scale job management in a metacomputing system. We investigate the use of a gateway model for controlling job placement on sub-clusters of a larger cluster of resources. Keywords: metacomputing; scheduling; cluster computing; adaptive scheduling. 1 Introduction The term metacomputer was coined by Fox [4] to d..