Faculty Advisor

Claypool, Mark L.

Abstract

Load sharing in a Beowulf cluster can be done transparently by PANTS. However, PANTS distributes load using only a measurement of CPU usage to index load. While CPU usage is the typical metric used in load distribution, other system resources such as disk and memory can become loaded and be a performance bottleneck. We examine PANTS in the context of load distribution algorithms, build new load indices, develop benchmarks, and evaluate performance. We find our load indices can reduce compile time of the Linux kernel by 1/2 of the time of the original PANTS indices.

Publisher

Worcester Polytechnic Institute

Date Accepted

January 2002

Major

Computer Science

Project Type

Major Qualifying Project

Accessibility

Restricted-WPI community only

Advisor Department

Computer Science

Share

COinS