Academic Commons

Reports

Methods for Performance Evaluation of Parallel Computer Systems

Eisenstadter, Yoram

Although parallel computers have existed for many years, recently there has been a surge of academic, industrial and governmental interest in parallel computing. Commercially manufactured parallel computers have started to become available. Many new experimental parallel architectures are reported in the literature every year. Software for many types of applications, from scientific number crunching to artificial intelligence, is being written to run on parallel machines. Performance is an essential consideration both in the design of new systems and the deployment of existing systems. Users of computers wish to utilize their hardware and software systems as efficiently as possible. Over the years, a field known as computer performance evaluation has arisen to address the problem of quantifying and predicting computer performance. Methods exist that can determine how efficiently a system's resources are being used. These can help track down the probable causes of performance problems.

Subjects

Files

More About This Work

Academic Units
Computer Science
Publisher
Department of Computer Science, Columbia University
Series
Columbia University Computer Science Technical Reports, CUCS-246-86
Published Here
November 2, 2011