Adaptive virtual machine consolidation framework based on performance-to-power ratio in cloud data centers

Abstract

Efficient resource management in a Cloud data center relies on minimizing energy consumption and utilizing physical resource efficiently while maintaining the service-level agreement (SLA) at its highest level. To achieve this goal, dynamically consolidating virtual machines (VMs) is considered a promising method, because it eliminates the hotspots resulting from overloaded hosts and switches the underloaded hosts to sleep mode through the live migration of VMs. However, during the consolidation, each VM migration consumes additional resource, leading to performance degradation and SLA violation. To address this issue, this study proposes a novel adaptive performance-to-power-ratio (PPR)-aware dynamic VM consolidation framework based on both the predicted resource utilization and PPR of the heterogeneous hosts to resolve the trade-off of performance and energy. The proposed framework consists of four stages: (1) host overload detection based on residual available computing capacity; (2) selection of the appropriate VMs for migration from the overloaded hosts based on minimum data transfer; (3) host underload detection based on multi-criteria Z-score approach; (4) allocating the VMs selected for migration from the overloaded and underloaded hosts based on the modified power-aware best-fit decreasing algorithm. To validate the reliability and scalability of the proposed method, we performed experimental evaluation in both real and simulated environments. The experimental results demonstrate that the proposed approach can reduce the energy consumption effectively and ensure maximal conformity to the quality of service (QoS) requirements across heterogeneous infrastructures, in comparison with the existing competitive approaches.