A parallel adaptive tabu search approach

This paper presents a new approach for parallel tabu search based on adapti ve parallelism. Adaptive parallelism was used to dynamically adjust the par allelism degree of the application with respect to the system load. Adaptiv e parallelism demonstrates that high-performance computing using a hundred of heterogeneous workstations combined with massively parallel machines is feasible to solve large optimization problems. The parallel tabu search alg orithm includes different tabu list sizes and new intensification/diversifi cation mechanisms. Encouraging results have been obtained in solving the qu adratic assignment problem. We have improved the best known solutions for s ome large real-world problems. (C) 1998 Elsevier Science B.V. All rights re served.