Energy-Efficient Load-Aware VM Placement using Multi-Metrics Analysis

Abstract

Objective: Cloud data centres comprise of a huge number of physical machines which host a number of heterogeneous virtual machines. These VMs exhibit fluctuating behaviour in terms of resource usage and specifications leading to disproportionate resource utilization in physical machines. This load imbalance on a host degrades its performance and violates many SLA features. To maintain even load in a data centre, an effective virtual machine placement technique is required. This paper proposes an efficient VM placement and migration technique based on three-tier architecture and considers load as a prime objective. Method: To maintain an even-load in a data centre, many load balancing techniques have been suggested which consider either migration cost or heterogeneity as the overall objective. This paper uses a multi-metrics analysis to distribute VMs evenly and maintains a stable equilibrium inside a data centre. We have applied Analytical Hierarchy Process (AHP) for efficient virtual machine placement using four post placement metrics which defines some of the key Service Level Agreement (SLA) parameters. The metrics considered for placement and migration are all load-centric and promise fewer migrations and SLA violations. Findings: Simulation results show a remarkable reduction in migrations which improves energy conservation inside the data centre. Application of AHP in balancing a data centre’s load is still unexplored. The presented placement technique selects the best candidate machine for placement, hence upgrades the performance. Improvement: Further enhancement includes adding more metrics for placement to fine tune performance and ensure better resource utilization along with reduction in overall energy consumption of a data centre.