Portland, Oregon
June 12, 2005
June 12, 2005
June 15, 2005
2153-5965
12
10.848.1 - 10.848.12
10.18260/1-2--14699
https://peer.asee.org/14699
648
Session 3620
Job Scheduling in Cluster Computing: A Student Project
Hassan Rajaei, Mohammad B. Dadfar
Department of Computer Science Bowling Green State University Bowling Green, Ohio 43403 Phone: (419)372-2337 Fax: (419)372-8061 {rajaei, dadfar}@cs.bgsu.edu
Abstract
Cluster computing has become widespread by utilizing COTS (Commercial-Off-The-Shelf) PCs, a high-speed network, and Linux operating system. This simple configuration of multiprocessor system can provide an excellent environment for student projects in courses such as Operating Systems, Data Communication, Distributed Programming, just to name a few. In this paper we describe an ongoing project focused on job scheduling for a cluster of processors.
Job scheduling on distributed-memory parallel systems has always been a challenge. Traditional measurement factors such as job length to allocate the requested resources does not suffice. Other factors such as communication delays and synchronization overhead which are normally in the user domain, could turn out as key issues for multiprocessors' utilization. As a result, utilization of each processor in a distributed-memory parallel system may end up comparatively lower than a single processor system. Consequently, performance of the entire system may degenerate and user jobs risk waiting long in the queue before getting the requested number of processors. This project is divided into several phases. In each phase, one to three students investigate how to minimize waiting time of the jobs in the queue while allowing other projects have a scheduling policy that suits their experiment and research. This paper reports the results of the first group which focuses on variable partitioning scheme. When resources for the highest priority job are not available, then the lower priority jobs are allowed to acquire the available resources. This paper investigates and evaluates variable partitioning schemes for job scheduling on distributed-memory parallel systems.
1. Introduction
High Performance Computing (HPC) nowadays can easily be achieved with clusters of PCs connected through a high-speed switch on a high-speed network. Such a tool provides excellent opportunities to explore numerous projects for educational as well as research purposes in computer science. For this reason, we have installed a Beowulf Cluster1, 2, 3, 4, 5 with 16 compute nodes in our computing lab in order to engage our students with exciting projects in courses such
Proceedings of the 2005 American Society for Engineering Education Annual Conference & Exposition Copyright © 2005, American Society for Engineering Education
Rajaei, H., & Dadfar, M. (2005, June), Job Scheduling In Cluster Computing: A Student Project Paper presented at 2005 Annual Conference, Portland, Oregon. 10.18260/1-2--14699
ASEE holds the copyright on this document. It may be read by the public free of charge. Authors may archive their work on personal websites or in institutional repositories with the following citation: © 2005 American Society for Engineering Education. Other scholars may excerpt or quote from these materials with the same citation. When excerpting or quoting from Conference Proceedings, authors should, in addition to noting the ASEE copyright, list all the original authors and their institutions and name the host city of the conference. - Last updated April 1, 2015