A DVFS Based Energy-Efficient Tasks Scheduling in a Data Center

With the popularity of cloud computing and high performance computing, the size and the amount of the datacenter develop rapidly, which also causes the serious challenges on energy consumption. Dynamic voltage and frequency scaling (DVFS) is an effective technique for energy saving. Many previous works addressed energy-officiate task scheduling based on DVFS. However, these works need to know the total workload (execution time) of tasks, which is difficult for some real-time tasks requests. In this paper, we propose a new task model that describes the QoS requirements of tasks with the minimum frequency. In addition, we define energy consumption ratio (ECR) to evaluate the efficiency of different frequencies under which to execute a take. Thus, it is possible to convert the energy-efficient task scheduling problem into minimizing the total ECR. By transforming the problem to the variable size bin packing, we prove that the minimization of ECR is NP-hard in this paper. Because of the difficulty of this problem, we propose task allocation and scheduling methods based on the feature of this problem. The proposed methods dispatch the coming tasks to the active servers by using servers as less as possible and adjust the execution frequencies of relative cores to save energy. When a task is finished, we propose a processor-level migration algorithm to reschedule remaining tasks among processors on an individual server and dynamically balance the workloads and lower the total ECR on this server. The experiments in the real test-bed system and simulation show that our strategy outperforms other ones, which verifies the good performance of our strategy on energy saving.