Data collection for global scheduling in the GENESIS system

Abstract

In order to schedule parallel processes on available computers there is a need to collect information about different cluster resources. Currently, global scheduling of parallel processes of applications takes into consideration mainly processor load and only in a small number of projects memory utilization. Communication costs, which are high in clusters, are practically neglected. The aim of this paper is to report on our study into cluster parameter measurements, in particular those, which characterize communication costs, the development of the resource discovery server and its initial testing through the study of the influence of communication costs on parallel application performance.