Performance Evaluation of an Experimental Grid Computer using MPI Applications

  • I. Ungurean Faculty of Electrical Engineering and Computer Science
  • S. G. Pentiuc Faculty of Electrical Engineering and Computer Science
  • V. Gaitan Faculty of Electrical Engineering and Computer Science

Abstract

A discussion about parallel programs performance is presented. These programs were developed using the OpenMPI implementation of the Message Passing Interface (MPI) standard and were run on an experimental grid computer made of 7 desktop computers. These test applications measure execution times for applications implementing the Jacobi approximation for a linear system of equations. The output of the test application is the processing speed, measured in MFlops and obtained using more methods in order to test more aspects of inter-process communication. The final purpose is to see increasing performance in a Grid by increasing the number of computing nodes. Ill. 8, bibl. 10 (in English; summaries in English, Russian and Lithuanian).

Published
2015-03-05
Section
T 120 SYSTEM ENGINEERING, COMPUTER TECHNOLOGY