Hybrid model for load balancing and server consolidation in cloud

Abstract

One of the most pressing and critical issues in cloud is Power Optimization. Due to the popularity of cloud, many computing applications are being hosted in the cloud. Naturally, many Cloud Centers (CCs) are experiencing huge power consumption problem, which leads to higher operational cost and environmental hazards due to carbon emissions. The two important concepts used to achieve power optimization are: load balancing and server consolidation. The first concept aims to achieve fair distribution of computing load on different Physical Machine (PM); whereas, the second concept aims to shutdown PMs, which are having limited computational load. One of the most popular techniques to achieve load balancing and server consolidation is Virtual Machine (VM) migration technique; where, the VMs are packed inside limited PMs; such that, load is fairly balanced among PMs, and limited usage PMs are relieved of their computational load and can be shutdown. However, VM migration requires excessive cost, and results in excessive computed task wastage. Hence, in this work, a new hybrid scheme which does not migrate VMs, and migrates only suitable tasks from overloaded VMs and lightly loaded PMs is presented. The proposed hybrid model is compared against contemporary VM migration technique to assess resource usage merits. The proposed hybrid model significantly outperforms the contemporary VM migration techniques in resource usage efficiency.