Energy-aware data center networks

With developing on cloud computing, data center networks have received great attention recently. In order to support high availability and bandwidth, data center networks are usually designed based on peak traffic. Unfortunately, peak traffic does not happen all the time. This causes unnecessary energy wastage and leads to low network utilization. We aim to minimize the power usage of data center networks by consolidating traffic flows and turning off unnecessary network devices. To support high availability, a requirement of two node-disjoint paths between the source and destination of each traffic flow is considered. We formulate the problem in mathematical format by integer linear programming (ILP). Due to its high computational complexity, we further propose a heuristic approach for practical use. The experimental results show that the heuristic approach reduces the power consumption by 15 to 35% in average while improving the availability by about 15% with significant improvement on computation time.