Dynamic Flow Scheduling for Power Optimization of Data Center Networks

In order to satisfy high bandwidth and availability requirements, data center networks (DCNs) are usually provisioned for peak workload. However, in the DCNs peak traffic does not always happen, which leads to tremendous energy wastage and low network utilization. Aiming to reduce the energy consumption in data center, we propose FLOWP, a dynamic adaptive scheduling algorithm based on flow preemption and power-aware routing, which saves energy by decreasing the ratio of low utilization devices and then put them into sleep mode. By the combination of the most energy saving path and set the tolerance threshold, FLOWP effectively optimized the energy consumption of the data center network. The experimental results show that on the premise of not affecting the quality of service (QoS), FLOWP could reduce energy by around 30% compared with ECMP algorithm.