Parallel Tabu Search for optimizing the OSPF weight setting problem

The problem of setting the Open Shortest Path First (OSPF) weights on links such that congestion can be avoided is proved to be NP-hard. Many iterative heuristics have been applied to solve the OSPF weight setting (OSPFWS) problem. As the size of any combinatorial optimization problem increases, it becomes more difficult to find an optimum solution using sequential algorithms. Parallelization of modern iterative heuristics has been proven to produce improved solution precision and timing. In this paper, we investigate the parallelization of Tabu Search and apply two variants of a Parallel Tabu Search (PTS) heuristic on the OSPFWS problem. It is shown through experimental results that both PTS approaches produced better solutions quality compared to the sequential heuristics; specifically for larger topologies. In one approach, we propose a new design for our parallel cooperative search algorithm, which performs better than the conventional parallel heuristic. The purpose of this new design is to induce diversification into the search to explore a larger search space. We also show that the new approach performs better than the conventional parallel heuristic.