Exact algorithms for energy-efficient virtual machine placement in data centers

Abstract

Virtual machine placement (VMP) and power management are essential topics in the development of cloud computing and data centers. The assignment of a virtual machine to physical machine impacts the energy consumption, the makespan, and the idle time of physical machines. In this paper, we formulate the problem as a three-dimension bin-packing optimization to minimize the energy cost of working machines and idle machines. By considering the CPU and memory requirements from a virtual machine, the assignment is constrained under the capacities of the physical machine. Inspired by the best-fit decreasing algorithm, four variants of this exact algorithm are developed to address the multiple-objective problem under multiple-capacity constraints. Experimental results demonstrate the effectiveness of the proposed algorithms on small-, medium- and large-scale instances profiled from data centers. The results indicate that the algorithms assigning virtual machines to the physical machines of best-fit hosting time is competitive in cases with loose capacity constraints, and the energy-efficiency best-fit algorithm produces efficient assignments when a makespan limit is required on the physical machines. The algorithm combining the fit rules has a linear computing time concerning the numbers of physical and virtual machines, and a stable performance that obtains gaps of results lower than 5.8% compared to an on-the-shelf mixed-integer linear program solver.