Optimizing the energy efficient VM placement by IEFWA and hybrid IEFWA/BBO algorithms

Abstract

In this paper, we present a problem-specific, information-based enhanced fireworks algorithm (IEFWA) and a hybrid of the IEFWA and the Biogeography-based optimization (BBO) algorithm. These new algorithms are tested for virtual machine (VM) placement problem with the objective of minimizing the energy consumption in datacenters, which is an integer space optimization problem. The EFWA algorithm is a relatively recent development in swarm intelligence (SI), and is based on explosion amplitude operator that operates in the continuous space. The 'round' function is used to convert the explosion amplitude value to the nearest integer to operate in integer space. In our IEFWA algorithm design, some domain knowledge of VM placement problem was used. During the spark generation in IEFWA, the explosion amplitude is added to the information-based selected components of the fireworks instead of adding the explosion amplitude into instead of all components. The hybrid IEFWA/BBO algorithm probabilistically chooses the explosion amplitude operator of IEFWA algorithm or the migration operator of BBO algorithm with a user determined probability for the exploitation of good candidate solutions. The VM placement problem is NP-hard, and existing results demonstrate that evolutionary algorithms (EAs) can be useful choice for good-quality solution with reasonable computing resources. We experimentally compare the performance of BBO, EFWA, IEFWA, hybrid IEFWA/BBO and the first fit decreasing (FFD) algorithms. Simulation results demonstrate the two key findings of this study. First, IEFWA algorithm consumes less CPU time as compared to the EFWA algorithm. Second, IEFWA and hybrid IEFWA/BBO algorithms outperform EFWA and BBO algorithms in terms of average energy consumed in datacenters.