VM Failure Prediction based Intelligent Resource Management Model for Cloud Environments

Abstract

This paper proposes a Virtual Machine (VM) failure prediction based intelligent cloud resource management (FP-IRM) model that estimates failure of VMs proactively and assorts all the available resources effectively. Specifically, a novel ensemble predictor is developed to determine any resource (CPU, storage) congestion prior to occurrence in real-time. Accordingly, the VM migration process is triggered proactively to proficiently manage the VM failures by reason of insufficient physical resources. FP-IRM model is implemented and evaluated by using a real-world benchmark Google Cluster VM traces dataset. The experimental simulation and comparison with state-of-the-arts confirms the influential performance of the proposed model which has reduced the number of active servers up to 51.2 % and an improved resource utilization up to 24.3 % over the comparative approaches.