Energy saving and load balancing for SDN based on multi-objective particle swarm optimization

Abstract

With the rapid development of cloud computing and large-scale data centers, the problem of network energy consumption is increasingly prominent. Most of the energy saving strategies on current IP network only aggregate traffic into a part of links. It leads to imbalance link utilization and seriously impacts the quality of service. With the emergence of the software defined network, the intelligent energy management becomes possible. In this paper, we take advantage of the centralized control and global vision of SDN to achieve the network energy saving and load balancing by dynamically aggregating and balancing of the traffic while ensuring QoS. We add actual QoS constrains to the basic maximum concurrent flow problem to formulate a multi-objective mixed integer programming model and we propose a multi-objective particle swarm optimization algorithm called MOPSO to solve this NP-hard problem. MOPSO distribute optimal paths for dynamic traffic demands and make idol switches and links into sleeping mode. Simulation results on real topologies and traffic demands show the effectiveness of our algorithm both on the objective of energy saving and load balancing compared with other algorithms.