Towards Energy Efficient Data Centers: A DVFS-Based Request Scheduling Perspective

Energy efficiency is a critical issue for cloud computing as more and more Internet services are deployed in data centers. It is observed that the energy consumption increases significantly as the CPU frequency gets higher. To reduce the total energy, Dynamic Voltage and Frequency Scaling (DVFS) technique is proposed to make the CPUs working at proper frequencies. However, lower frequency will reduce the computing capacity of the servers, which leads to using more servers. It is a challenging problem to make a tradeoff between the number of servers and the frequency of each server for a given workload. In this paper, we prove that the problem of dynamic resource allocation based on DVFS with the target of minimizing energy consumption is NP-Hard. And we propose two algorithms based on different basic ideas to solve this problem. We also compare our algorithms to the well-known First Fit Decreasing (FFD) algorithm, and the simulation results show that we can get 12%-14% power saving on average although we use 1.2x-1.3x number of servers compared to FFD.