A Multi-objective Ant Colony Algorithm for Deployment Optimization of Internetware Application

Changes in operating environment may result in the performance degradation and cost overruns to an Internetware application. An efficient way to solve such problems is to optimize its deployment architecture according to the changes. However, there may be many different deployment architectures and the optimal ones should exhibit right trade-offs among conflicting objectives. Finding optimal deployment architectures for an Internetware application is hard and time consuming. This paper propose to employ a multi-objective ant colony algorithm MACO-DO to explore the search space automatically, aiming at finding a set of pareto optimal deployment architectures for an Internetware application. This algorithm is an improved version of traditional algorithms. It introduces a discarding elitist strategy to prevent algorithm from premature convergence. A series of experiments are implemented on three simulated instances of different sizes to compare the proposed MACO-DO with recently proposed P-ACO and NSGA-II. The results show that MACO-DO has better performance than others on the considered problem.