An Ant Colony System for energy-efficient dynamic Virtual Machine Placement in data centers. (15th April 2019)

• Record Type: Journal Article
• Authors: Alharbi, Fares; Tian, Yu-Chu; Tang, Maolin; Zhang, Wei-Zhe; Peng, Chen; Fei, Minrui
• Is Part Of: Expert systems with applications. Volume 120(2019)
• Page Start: 228
• Page End: 238
• Publication Date: 2019-04-15
• Subjects: Data center -- Energy efficiency -- Optimization -- Virtual Machine Placement -- Profile -- Ant Colony System; Expert systems (Computer science) -- Periodicals; Systèmes experts (Informatique) -- Périodiques; Electronic journals; 006.33
• DOI: 10.1016/j.eswa.2018.11.029
• Languages: English
• ISSNs: 0957-4174

Abstract

Highlights: Dealing with dynamic virtual machine placement in data centers for energy efficiency. Formulated the problem as a constrained optimization with profile information. An ant colony system embedded with new heuristics to solve the problem. Significant increase in energy efficiency of data centers. Abstract: Data centers are fundamental infrastructure for information technology and cloud services; however, their very high rates of energy consumption are a problem. The placement of Virtual Machines (VMs) to Physical Machines (PMs) in virtualized environments has a significant impact on the energy consumption of a data center. This is an NP-hard problem, for which an optimal solution is not practicable even for a small-scale data center. In this paper, we formulate placement of VMs to PMs in a data center as a constrained combinatorial optimization problem and make use of the information from PM and VM profiles to minimize the total energy consumption of all active PMs. An Ant Colony System (ACS) embedded with new heuristics is presented for an energy-efficient solution to the optimization problem. To demonstrate the effectiveness of the ACS, simulation experiments are conducted on small-, medium- and large-scale data centers. The results from our ACS are compared with two existing ACS methods as well as the widely used First-Fit-Decreasing (FFD) algorithm. Our ACS is shown to outperform the two existing ACS methods and FFD in energy performance for all small-, medium- and large-scale test problems. Our ACS also exhibits good scalability with the increase in the problem size.