Energy-Efficient Load Balancing in a SDN-based Data-Center Network

The Software Defined Networking (SDN) paradigm brings flexibility in the network management and can be used in order to reduce the energy consumption of the Data Center (DC) networks. In particular, two main leverages can be exploited to reduce brown energy consumption: a sleep mode on hosts in the DCs and geographical load balancing of the requests. In this paper, we propose a mixed integer linear programming formulation to compute the optimal requests assignment to data centers, with both a multi-period approach and a period-by-period assignment. We evaluate the impact of knowing future requests with respect to the optimal assignment. In addition, we provide an efficient on-line algorithm that could be implemented in an operational setting. The evaluation of the algorithm is based on real traffic traces, and shows a reduction up to 42% in the brown energy consumption.