A prediction and fuzzy-based approach for VM migration for optimal trade off energy-performance in cloud network

Abstract

Abstract Energy efficiency, low utilization of Physical Machines (PM) and service access level are challenges of cloud networks, while efficient migration of Virtual Machines (VMs) is a solution for them. Uncontrolled migration of the VMs can causes migration overhead, that most important of them comprising: violation in Service Level Agreement (SLA) and high energy consumption. Generally, performance requirements in the form of SLAs are formalized. The main goal of the dynamic migration of VMs is a trade-off in energy and performance. Detecting when cloud's PMs is being overloaded or under-loaded are two substantial sub-problems of dynamic VM migration, which directly effects the performance and energy efficiency. Thus, migration technique is a practical approach for balancing the electricity consumption and Quality of Service (QoS) requirements in the cloud networks. In this paper, with definition an architectural framework in the system modelling, in first to detect overloaded and under-loaded PMs a Predict Detection method (PD method), and then for selection of VMs on overloaded PMs a Fuzzy Selection method (FS method), in overall, algorithm (PDFS) is proposed. The algorithm PDFS simulation results demonstrate that the proposed algorithm significantly outperforms benchmark algorithms in terms of energy, migration overhead, service access and SLA violation.