Home About Research Invention Members area Contact
Resources
Projects
Publications
Presentations
Supports

ACM Computing Reviews

Grid harvest service: a performance system of grid computing
Wu M., Sun X. Journal of Parallel and Distributed Computing 66(10):
1322-1337, 2006. Type: Article
Date Reviewed: Apr 5 2007

Wu and Sun describe the design and implementation of grid harvest service (GHS), a system for performance evaluation and task scheduling of large-scale applications in shared environments. GHS is specifically for long-term application-level performance prediction, in contrast to the existing systems that are designed for short-term applications. The components of GHS are a performance monitor, a performance predictor, a task partitioner and scheduler, and a user interface; the functions of these are presented in detail.

A new model for performance evaluation implemented in GHS is introduced. Instead of the classical use of a fixed utilization for resource availability, the new model takes into account resource use. The model identifies the effects of machine utilization, computing power, local job service, and task allocation on application performance. The authors pay special attention to the prediction of application execution time and the prediction of resource availability.

Another issue that is carefully analyzed is task scheduling. The GHS task scheduler supports different scheduling algorithms for different application requirements; for example, a self-adaptive task-scheduling algorithm is designed and implemented to deal with a situation where resources present abnormal performance from their historical records. Finally, the experiments performed on real environments and applications highlight the benefits of using GHS and the correctness of its models.

Another useful aspect of the paper is the state-of-the-art presentation concerning performance evaluation and prediction models and tools. The design of the software prototype and the experimentation results are comprehensive. Therefore, the paper can be considered a guide to GHS.

Reviewer: Petcu Dana Review #: CR134109




Illinois Institute of Technology
Home | About | Contact | Sitemap
Copyright 1999-2006. All rights reserved. Last updated on April 18, 2007