An improved genetic algorithm for efficient scheduling on distributed memory parallel systems

A key issue related to the distributed memory multiprocessors architecture for achieving high performance computing is the efficient scheduling of heavily communicated parallel applications such that the total execution time is minimized. Therefore, this paper provides a genetic algorithm based on task clustering techniques for scheduling parallel applications with large communication delays on distributed memory parallel systems. The genetic algorithm is improved with the introduction of some extra knowledge about the scheduling problem. This knowledge is represented by a class of clustering heuristic which is based on structural properties of the parallel application. The major feature of the proposed algorithm is that it takes advantage of the effectiveness of task clustering for reducing communication delays combined with the ability of the genetic algorithms for exploring and exploiting information of the search space of the scheduling problem. The algorithm is assessed by simulation run on some families of traced graphs which represents some of the numerical parallel application programs, and a set of randomly generated applications. Simulation results showed that this algorithm significantly improves the performance of related approaches.