HBWO-CML: An energy-efficient Hybrid Black Widow Optimisation and a Combined Machine Learning approach for VM Placement

Abstract

Cloud data centers remove the hardware barriers for the world with enormous storage and computing capabilities and provide access from anywhere over the internet with the concept of virtual machines (VM) and physical machines (PM). These cloud centers raise the problem of lack of resource utilisation and energy waste. These drawbacks have led many researchers to propose VM placement techniques that improve resource utilization and energy consumption. The research proposes VM migration/consolidation and host selection techniques to improve the resource utilisation of cloud centers. This research suggests hybrid black widow optimization (HBWO) and a combined machine learning model to mitigate the problem of cloud centers. HBWO optimizes the VM placement sub-process using an evolutionary approach and a combined machine-learning-based model to predict the over and under-utilised hosts. The effectiveness of the proposed approach was analysed and compared with a Power-Aware Best-Fit Decrease (PABFD) approach and a Particle Swarm Optimization (PSO) approach. The proposed approach improves energy utilisation by 30-40\% over PABFD and 20-15\% over PSO. Moreover, the combined approach of HBWO and machine learning improves the overall performance by 40-50\% compared to PABFD and 20-30\% compared to PSO.