Multi-core Deployment Optimization Using Simulated Annealing and Ant Colony Optimization

摘要

This work introduces a hybrid metaheuristic algorithm for solving the problem of multi-core deployment optimization (MCDO). It extends prior work using Ant Colony Optimization to solve MCDO by initially seeding the pheromone matrix with the output of a Simulated Annealing metaheuristic. This work also removes a number of critical simplifying assumptions from the MCDO model. Across 28, 800 different algorithm inputs, the hybridized algorithm is shown to have a median improvement in makespan time of 7.2% versus the nonhybrid version, as well as a median reduction of 74% in execution time. On a dataset of 50 MCDO problems with known optimal solutions, the median hybrid algorithm solution is 16.5% worse than known optimal.