A novel framework for the parallel solution of combinatorial problems implementing tabu search and simulated annealing algorithms

In this paper, we propose a novel framework for the parallel solution of combinatorial problems. The main idea behind this approach is to explore sub-domains of the solution space making use of different metaheuristics. This is developed in three basic steps: a main computing node sends an initial condition to his workers. The workers make use of tabu search and simulated annealing algorithms and submit its optimal solution to the master node; then, the master selects the best solution among all the workers and returns them the best solution as new initial condition. The novelty of this framework is that the relevance of the solution is not in the number of iterations but in the increment of the solution space by increasing the number of instances. Experimental settings are carried out making use of 16, 32, 64, 128, 256, 512, 1024 and 2048 processors (workers) in the Blueridge super computer at Virginia Tech, USA. Instances of the TSP-LIB (Traveling Salesman problem) were chosen in order to perform the tests. The results reveal that when the number of processors increases, the optimal solutions in the global optimization process are considerably improved.