An Intelligent and Adaptive Threshold-Based Schema for Energy and Performance Efficient Dynamic VM Consolidation

Dynamic VM consolidation as a dynamic control procedure is an effective way to improve energy efficiency in cloud data center. In general, the problem of dynamic VM consolidation can be split into four sub-problems (1) host overloading detection; (2) host under-loading detection; (3) VM selection and (4) VM placement. The goal of this procedure is to operate in a way that optimizes energy-performance tradeoff inside cloud data center. To this end, each of aforementioned sub-problems must operate in an optimized manner. In this paper with concentration on host overloading detection sub-problem, we propose an intelligent and adaptive threshold-based algorithm for detecting overloaded hosts by Dynamic Fuzzy Q- learning (DFQL). Our algorithm with interaction of dynamic characteristics of each physical host environment learns when a host must be considered as an overloaded host towards energy-performance tradeoff optimization inside data center. We validated our approach with the CloudSim toolkit using real world Planetlab workload. We show that our approach outperforms the state of the art algorithms.