A software tool for accurate estimation of parameters of heterogeneous communication models

Abstract

Analytical communication performance models play an important role in prediction of the execution time of parallel applications on computational clusters, especially on heterogeneous ones. Accurate estimation of the parameters of the models designed for heterogeneous clusters is a particularly challenging task due to the large number of parameters. In this paper, we present a set of communication experiments that allows us to get the accurate estimation of the parameters with minimal total execution time, and software that implements this solution. The experiments on heterogeneous cluster demonstrate the accuracy and efficiency of the proposed solution. © 2008 Springer-Verlag Berlin Heidelberg.