Open Access Open Access  Restricted Access Subscription Access
Open Access Open Access Open Access  Restricted Access Restricted Access Subscription Access

Comparative Study of Performance Metrics for Evaluating Parallel Computing Systems


Affiliations
1 Deptt. of Comp. Sci. & App., DAV College, Abohar (Punjab)., India
     

   Subscribe/Renew Journal


A sequential algorithm is usually evaluated in terms of its execution time which is expressed as a function of its input size. On the other hand, the execution time of a parallel algorithm depends not only on the input size but also on the parallel architecture and the number of processors employed. Hence, a parallel algorithm cannot be evaluated isolated from a parallel computer. A parallel system should be viewed as a combination of a parallel algorithm and the parallel computer on which it is implemented. This paper presents a review of some performance metrics, standard measures, and benchmark for evaluating the performance of a parallel computing system.

Keywords

Parallel Systems, Performance Metrics, Performance Evaluation, Speedup, Efficiency, Superlinear Speedup, Cost
Subscription Login to verify subscription
User
Notifications
Font Size


  • V.Rajaraman & C.Siva Ram Murthy “Parallel Computers Architecture and programming”, Prentice Hall.
  • http://www.cc.gatech.edu/~rama/TASS/papers/git.cc.94.38.pdf.
  • Introduction to Parallel Computing, Ananth Grama, et. al., Pearson Education, Ltd., Second Edition, 2004.
  • Sun and Gustafson. Towards a better parallel performance metric. Parallel Computing.
  • Sun X-H, and Ni, l. Scalable problems and memory bounded speedup.

Abstract Views: 313

PDF Views: 2




  • Comparative Study of Performance Metrics for Evaluating Parallel Computing Systems

Abstract Views: 313  |  PDF Views: 2

Authors

Sanjeev Kumar
Deptt. of Comp. Sci. & App., DAV College, Abohar (Punjab)., India

Abstract


A sequential algorithm is usually evaluated in terms of its execution time which is expressed as a function of its input size. On the other hand, the execution time of a parallel algorithm depends not only on the input size but also on the parallel architecture and the number of processors employed. Hence, a parallel algorithm cannot be evaluated isolated from a parallel computer. A parallel system should be viewed as a combination of a parallel algorithm and the parallel computer on which it is implemented. This paper presents a review of some performance metrics, standard measures, and benchmark for evaluating the performance of a parallel computing system.

Keywords


Parallel Systems, Performance Metrics, Performance Evaluation, Speedup, Efficiency, Superlinear Speedup, Cost

References