Network power saving based on Pareto optimal control with evolutionary approach

The power consumption of networks has been increasing as the service over the Internet becomes popular, and has become a serious problem. Many methods to reduce the power consumption by shutting down unnecessary network devices following the environmental changes have been proposed. These methods consider only simple objectives such as the number of powered-on nodes and the maximum link utilization. However, multiple complex objectives such as delay and reliability should be also considered in the actual network. In this paper, we propose a network power saving method that handles multiple complex objectives, following the environmental changes. In this method, we store the candidate network configurations, and evolve them, following the environmental changes. Then, we select the network configuration from the candidate network configurations.We combine two approaches to evolve the network configurations. The first approach is based on Pareto optimal, and evolves the network configurations so as to be close to the Pareto optimal solutions, considering multiple objectives. Another approach is based on the diversity of the network configurations. By storing the diverse network configurations, we can handle the significant environmental changes. We evaluate our method by simulation, and demonstrate that our method reduces the power consumption without violating the constraints, following the traffic changes. In addition, we also demonstrate that our method can keep the connectivity in case of failures, and recover the performance and the small power consumption soon after the failure occurs.