Energy-aware parameter tuning for mixed workloads in cloud server

作者：Liang, Jiechao; Lin, Weiwei^*; Xu, Yangguang^*; Liu, Yubin; Mo, Ruichao; Luo, Xiaoxuan

来源：Cluster Computing-The Journal of Networks Software Tools and Applications, 2023.

DOI：10.1007/s10586-023-04212-6

• Cloud server
• Energy efficiency
• Mixed workloads
• Parameter tuning

Server energy consumption constitutes a significant portion of the overall energy usage in cloud data centers. Achieving energy optimization through tuning server parameter configurations is of paramount importance for energy conservation within cloud data centers. However, due to the diverse and dynamic nature of mixed workloads in cloud servers, achieving highly energy efficient server parameter configurations often necessitates continuous adjustments and optimizations in response to workload fluctuations. To tackle this challenge, this paper introduces an energy-efficient optimization method for cloud servers tailored to scenarios with mixed workloads, known as Energy-aware Parameter Tuning for Mixed Workloads (EPTMW). This method dynamically performs joint tuning of CPU frequency and system kernel parameters based on the operational status of the real-time workloads. Additionally, we evaluate EPTMW using benchmark workloads from BenchSEE and SERT. The experimental results demonstrate that EPTMW can enhance energy efficiency by 28.5%\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\%$$\end{document} compared to the energy efficiency level achieved under the default server parameter configuration, all while maintaining server performance.