AIRCC's International Journal of Computer Science and Information Technology

Srirangam V. Addepallil1High Performance Computing Centre, Texas Tech University, Lubbock, Texas ¹, Per Andersen ², George L. Barnes ³

Affiliations
1 High Performance Computing Centre, Texas Tech University, Lubbock, Texas, US
2 Department of Computer Science, Texas Tech University, Lubbock, Texas, US
3 Department of Chemistry, University of Oregon, Eugen, Oregon, US

Abstract

In this paper, a method for efficient scheduling to obtain optimum job throughput in a distributed campus grid environment is presented; Traditional job schedulers determine job scheduling using user and job resource attributes. User attributes are related to current usage, historical usage, user priority and project access. Job resource attributes mainly comprise of soft requirements (compilers, libraries) and hard requirements like memory, storage and interconnect. A job scheduler dispatches jobs to a resource if a job's hard and soft requirements are met by a resource. In current scenario during execution of a job, if a resource becomes unavailable, schedulers are presented with limited options, namely re-queuing job or migrating job to a different resource. Both options are expensive in terms of data and compute time. These situations can be avoided, if the often ignored factor, availability time of a resource in a grid environment is considered. We propose resource rank approach, in which jobs are dispatched to a resource which has the highest rank among all resources that match the job's requirement. The results show that our approach can increase throughput of many serial / monolithic jobs.

Keywords

SGE, LSF, Venus, ENDyne, Condor.

Full Text