Parallel Graph-Partitioning Using the Mob Heuristic

In this paper we study graph partitioning which is an important NP-complete problem. We have implemented two mob heuristics and analysed their performance. Our main result is that for randomly generated graphs the mob heuristics with an exponential schedule is more efficient than that with the linear schedule, proposed earlier. With the other parameters of the heuristic being identical the exponential schedule doubles the speedup.