An Energy Efficient VM Management Scheme with Power-Law Characteristic in Video Streaming Data Centers
Hsueh‐Wen Tseng, Ting-Ting Yang, Kai-Cheng Yang, Pei-Shan Chen · IEEE Transactions on Parallel and Distributed Systems · 2017
As cloud computing services have gained popularity, users view videos on websites (e.g., YouTube) to generate high CPU resource utilization and bandwidth for video streaming data centers. However, popular videos result in power-law features to cause imbalanced resource utilization. In addition, hotspot and idle servers generate extra power consumption in data centers. Previous studies considered to satisfy the requirements of users, provide faster access rates and save power consumption. However, fewer studies considered resource utilization with different popularity videos. Therefore, this paper proposes an energy efficient virtual machine (VM) management scheme with power-law features (VMPL). VMPL predicts the resource utilization of the video in the future based on the popularity, ensures enough resources for upcoming videos, and turns off idle servers for power saving. Simulation results validated by mathematical analysis show that VMPL has the best resource utilization and the lowest power consumption compared with Nash and Best-Fit algorithms.