An Approach for Detection of Overloaded Host to Consolidate Workload in Cloud Datacenter

Abstract

This article describes the process of workload consolidation through detection of overloaded hosts in Cloud datacenter which leads to saving in energy consumption. Cloud computing is a novice paradigm where virtual resources are provisioned on pay-as-you-go basis. Upon receiving users' job requirement, it is mapped onto virtual resources running on hosts in datacenter. To achieve workload consolidation, it is required to detect the overloaded hosts. Overloaded host detection is carried out for balancing workload, creating a list of overloaded hosts which will be useful while placing VMs (by not putting a VM on already overloaded host) to reduce Service Level Agreement (SLA) violation and while checking the underloaded host, the overloaded hosts are omitted to reduce computational cost. Most common mechanism to detect overloaded hosts is to calculate upper threshold values based on hosts' utilization statically or dynamically. Most researchers recommend dynamic calculation of threshold values. In this research, the authors propose to use moving range (MR) method of variables control charts to calculate upper threshold. The experimentation results show that MR performs better in terms of reduction in SLA violation, minimization in VM migration.