A heuristic approach to task assignment optimization in distributed systems

This paper describes an efficient heuristic algorithm for the task assignment problem (an NP-complete problem). The problem is to find an optimal mapping of multiple communicating tasks onto the processing nodes of a distributed computing system. The purpose of mapping these tasks onto the nodes of the system is the minimization of the total execution time without sacrificing solution quality. Many heuristic approaches have been employed to obtain satisfactory mapping. Our heuristic, the genetic mean field annealing (GMFA), is a hybrid of genetic algorithm and mean field annealing. The hybrid algorithm combines the benefit of both methods and thus improves the performance. We compare the quality of solutions and time derived by the proposed GMFA against those derived by the genetic algorithm and mean held annealing algorithm respectively. Our experimental results from a simulation study of the heuristic algorithms are presented. In all cases studied, the solution quality derived by our new approach were significantly better than those derived by the conventional approaches.