An Enhanced VM Selection and Migration Algorithm for Cloud Computing

Abstract

Virtual Machine (VM) allocation and migration has always been an area of interest for the researchers due to its high performing capabilities with an attempt of minimizing the total amount of power consumed in order to compensate the user supplied workload. The migration becomes essential to support elasticity in the performance network. The migration policy has four major factors namely the hotspot detection, selection of VMs from the hotspot, detection of the target Physical Machine (PM) and the migration. Different approaches have been presented till date to reduce the overall consumed power with least number of violations as per the Service Level Agreement (SLA). This research proposal extends the concept of VM selection and migration from hotspot PM to target PM. The research proposal is based on Nashaat et al. research proposal which extends K-means algorithm. Nashaat s work was based on CPU Utilization and associated RAM to choose total number of clusters to group similar kind of VMs so that the migration cost could be reduced. The proposed algorithm Cosine Elastic Clustered Scheduling Algorithm (CESCA) aims to enhance the existing Smart Elastic Scheduling Algorithm (SESA) by introducing Bandwidth Utilization (BU) in the existing selection policy. In addition to that, the concept of clustering would also be aimed to be enhanced by introducing similarity indexes in combination with Euclidian distance. To design a relocation policy, the proposed work uses T-MBFD for the reallocation of the selected VMs from over utilized PMs and underutilized PMs categorized as the hotspot PM. The reallocation policy is statistical in nature and it computes the best possible PM relative to the time. In order to automate the architecture, the proposed work uses Machine Learning (ML) based reputation system in which the proposed work divide the entire data into three possible groups viz. High Rank (HR) , Moderate Rank (MR) , and Low Rank (LR) . In order to divide the data into multiple groups, the proposed work utilized 3 c