Multi-objective Service Monitoring Rate Optimization using Memetic Algorithm

In dynamic service-oriented environment, service monitoring could provide reliability improvement to service composition as well as cost increase. To reduce the overall cost brought by monitoring, existing literatures proposed to decrease the number of monitors through monitoring the most reliability-sensitive services. However, the optimal monitoring rate for those monitors was not taken into account at the same time. Aiming at choosing optimal monitoring rate for minimal number of monitors, this paper proposed to search appropriate monitoring rate to minimize multi kinds of resources cost by monitoring under reliability constraints. Firstly, two multi-objective optimization problems were presented with the reliability and cost models of service composition under monitoring analyzed through Markov chain. Then a multi-objective memetic algorithm (MOMA) was used to search the near-optimal solutions of monitoring rate for services. This algorithm employed nondominated sorting strategy as the global search method and used random walk with direction exploitation method as local search operator. Experimental studies results showed that multi-objective approach for service monitoring rate optimization could provide solutions with a variety of trade-offs between the system reliability and cost comparing with existing greedy sensitivity-based method. Comparison with other multi-objective evolutionary algorithms showed that, in terms of both the coverage rate and hypervolume indicator, MOMA searched more effectively than several state-of-art algorithms including NSGA II, PHC-NSGA-II and HaD-MOEA.