Optimal task assignment in heterogeneous distributed computing systems

exploit effective parallelism on a distributed system, tasks must be properly allocated to the processors. This problem, task assignment, is well-known to be NPhard in most cases.1 A task-assignment algorithm seeks an assignment that optimizes a certain cost function—for example, maximum throughput or minimum turnaround time. However, most reported algorithms yield suboptimal solutions. In general, optimal solutions can be found through an exhaustive search, but because there are nm ways in which m tasks can be assigned to n processors, an exhaustive search is often not possible. Thus, optimal-solution algorithms exist only for restricted cases or very small problems. The other possibility is to use an informed search to reduce the state space. The A* algorithm, an informed-search algorithm, guarantees an optimal solution, but doesn’t work for large problems because of its high time and space complexity. Thus, we require a further-reduced state space, a faster search process, or both.