Scheduling of Parallel Jobs on Dynamic, Heterogenous Networks

In using a shared network of workstations for parallel pr ocessing, it is not only impor- tant to consider heterogeneity and differences in processing power between the worksta- tions but also the dynamics of the system as a whole. In such a computing envir onment where the use of r esources vary as other applications consume and r elease resources, intelligent scheduling of the parallel jobs onto the available r esources is essential to max- imize resource utilization. Despite this realization, however, there are few systems avail- able that provide an infrastructure for the easy development and testing of these intelligent schedulers. In this paper, an infrastructure is presented together with a scheduler that is capable of both gang scheduling and dynamic task r eallocation of PVM applications.