Improving the Scalability of Communication-Aware Task Mapping Techniques

The advent of cluster computing introduced some years ago the need for taking into account the communications that take place on distributed computer architectures when executing applications. In that environment, different communication-aware mapping techniques were proposed for improving the system performance, both for off-chip and for on-chip networks. Some of these proposals are based on heuristic search for finding pseudo-optimal assignments of a given population of tasks and processing elements. However, the technology has allowed a significant increase in the problem size, arising the scalability problem. In this paper, we propose the use of an evolutionary computation framework to implement a genetic algorithm that can significantly improve the scalability of communication-aware task mapping techniques. We have studied different genetic operators and selection mechanisms, choosing those providing the best performance for this particular problem. The performance evaluation results shows that for medium and large domain spaces, the genetic algorithm provides better solutions while requiring lower or similar execution times. These results indicate that the heuristic search based on genetic algorithms can improve the scalability of communication-aware task mapping techniques for both cluster computing and Networks-on-Chip.