Joint Energy Optimization of Cooling Systems and Virtual Machine Consolidation in Data Centers

• Main Authors: Liu, Hai, Wong, Wai Kit, Ye, Shujin, Yu Tak Ma, Chris
• Format: Conference Proceeding
• Language: English
• Subjects: cloud; Cloud computing; Cooling; Cooling system; data center; Data centers; Energy consumption; energy optimization; Memory management; Temperature distribution; VM consolidation
• ISSN: 2637-9430
• DOI: 10.1109/ICCCN49398.2020.9209712

Minimizing energy consumption of data centers is important to reduce carbon emissions. Virtual machines (VMs) consolidation is a typical technique to utilize the available data center resources and thus improve energy efficiency. The cooling systems consume up to 50% of the total data center electricity. In this work, we investigate the joint energy optimization of cooling systems and VM consolidations in cloud data centers. We propose a cooling-aware VM consolidation (CAVC for short) algorithm to the problem. The CAVC algorithm is a two-stage solution: 1) we first relax the constraints of the problem and determine an optimal number of physical machines (PMs) and an optimal CPU utilization of the PMs that yields the minimum cooling power; and 2) based on the initial solution of the first stage, we consolidate the VMs into the predetermined PMs with the predetermined CPU utilization ratio as much as possible. To the best of authors' knowledge, this is the first work that jointly considers the VM consolidation and the cooling systems in minimizing energy consumption of cloud data centers. We derive an approximation ratio of CAVC over the optimal solution. The real-world data set (i.e., Google cluster data) is adopted in the simulations and the results show that the CAVC algorithm yields very close energy consumption to the theoretical lower bound.