(IJACSA) International Journal of Advanced Computer Science and Applications, Vol. 7, No. 9, 2016 Optimization of Dynamic Virtual Machine Consolidation in Cloud Computing Data Centers Alireza Najari 1,* Department of Computer Engineering Ahvaz Branch, Islamic Azad University Ahvaz, Iran Department of Computer Engineering Khouzestan Science and Research Branch Islamic Azad University Ahvaz, Iran Seyed EnayatOllah Alavi 2 Department of Computer Engineering Shahid Chamran University of Ahvaz Ahvaz, Iran Mohammad Reza Noorimehr 3 Department of Computer Engineering Ahvaz Branch, Islamic Azad University Ahvaz, Iran Abstract—The present study aims at recognizing the problem of dynamic virtual machine (VM) Consolidation using virtualization, live migration of VMs from underloaded and overloaded hosts and switching idle nodes to the sleep mode as a very effective approach for utilizing resources and accessing energy efficient cloud computing data centres. The challenge in the present study is to reduce energy consumption thus guarantee Service Level Agreement (SLA) at its highest level. The proposed algorithm predicts CPU utilization in near future using Time-Series method as well as Simple Exponential Smoothing (SES) technique, and takes appropriate action based on the current and predicted CPU utilization and comparison of their values with the dynamic upper and lower thresholds. The four phases in this algorithm include identification of overloaded hosts, identification of underloaded hosts, selection of VMs for migration and identification of appropriate hosts as the migration destination. The study proposes solutions along with dynamic upper and lower thresholds in regard with the first two phases. By comparing current and predicted CPU utilizations with these thresholds, overloaded and underloaded hosts are accurately identified to let migration happen only from the hosts which are currently as well as in near future overloaded and underloaded. The authors have used Maximum Correlation (MC) VM selection policy in the third phase, and attempted in phase four such that hosts with moderate loads, i.e. not overloaded hosts, liable to overloading and underloaded, are selected as the migration destination. The simulation results from the Clouds framework demonstrate an average reduction of 83.25, 25.23 percent and 61.1 in the number of VM migrations, energy consumption and SLA violations (SLAV), respectively. Keywords—Cloud Computing; Dynamic Consolidation; Energy Consumption; Virtualization; Service Level Agreement I. INTRODUCTION According to the definition provided by NIST [1] "cloud computing is a model for enabling ubiquitous, convenient, on- demand network access to a shared pool of configurable computing resources (e.g., networks, servers, storage, applications and services). It can be rapidly provisioned and released with minimal management effort or service provider interaction". Infrastructure as a Service (IaaS) is among the services provided by cloud computing that offers processing resources to users as services. The clients rent the equipment from infrastructure providers as a service and only pay for the amount of service they really consume [2, 3]. The ever-increasing growth and wide applications of cloud computing, as well as the extensive usage of cloud services in all scopes, have caused a growing trend in energy consumption of cloud computing data centres. Therefore, the operational costs in these centres are intensively increasing due to the electric energy used [4]. According to the reports published by Microsoft [5], the consumed energy used by physical resources can account for 45 percent of the operational costs in a data centre. This amount has multiplied in the last five years [6]. Therefore, to maintain their business and to remain in the market competition, service providers need to minimize energy consumption to cut the excessive operational costs in a way that the integrity and quality of service remain intact [7]. Hence, two challenging tasks in IaaS are management and optimized allocation of resources, to the extent that the success of cloud services heavily relies on this issue. The present study recognizes the problem of dynamic virtual machine Consolidation using virtualization, live migration of VMs from underloaded and overloaded hosts and switching idle nodes to the sleep mode, as a very effective approach for utilizing resources and accessing energy efficient cloud computing data centres [8-11]. The challenges faced are the consolidation of VMs and their allocation and placement on physical service providers in a way that they minimized energy consumption in the entire data centre and the number of active hosts as well as SLA violation, which is a contract between the clients and the providers. Many studies [11-13] have reported that fully idle 202 | Page www.ijacsa.thesai.org