A decrease-and-conquer genetic algorithm for energy efficient virtual machine placement in data centers

• Published in: 2017 IEEE 15th International Conference on Industrial Informatics (INDIN) pp. 135 - 140
• Main Authors: Sonklin, Chanipa, Maolin Tang, Yu-Chu Tian
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01-07-2017
• Subjects: Biological cells; Cloud Computing; Data Center; Dynamic Optimization; Energy consumption; Genetic Algorithm; Genetic algorithms; Optimization; Server Consolidation; Servers; Virtual Machine Placement Optimization; Virtual machining
• ISSN: 2378-363X
• DOI: 10.1109/INDIN.2017.8104760

The dramatically increasing energy consumption of data centers is an important issue and one of the most efficient ways to tackle the issue is through server consolidation. The basic idea of server consolidation is to move all virtual machines (VMs) to as few energy efficient servers as possible, and then switch off unused servers. Many efficient server consolidation approaches have been proposed and one of the most efficient approaches is to use a Genetic Algorithm (GA) to find an optimal or near-optimal solution to the server consolidation problem. Aiming at reducing the computation time and the number of VM migrations incurred by server consolidation, this paper proposes a Decrease- and-Conquer Genetic Algorithm (DCGA). This DCGA adopts a decrease-and-conquer strategy to decrease the problem size and to decrease the number of VM migrations without significantly compromising the quality of solutions. The DCGA is compared with a classical GA and the most popular approach, namely FFD, for the server consolidation problem by experiments and the experimental results show that the DCGA can find a solution very close to the solution generated by the classical GA with much shorter computation time and incur much less VM migrations for all the test problems, and that the DCGA can generate a much better solution than the FFD.