Predictive Virtual Machine Placement for Energy Efficient Scalable Resource Provisioning in Modern Data Centers

Abstract

In modern Data Centers (DCs), the major significant and challengeable task is resource management of cloud and efficient allocation of Virtual Machines (VMs) or containers in Physical Machines (PMs). There are several schemes proposed regarding this factor that includes VM placement considering utilization of resources. The process of consolidation may be done efficiently using “opportunities” discovery for migrating VMs and estimating utilization of resource to VM placement. However, the deduction of energy utilized over cloud DCs by physical resources with heterogeneous mode gets accomplished using consolidation of VM. This assists in minimize of PM numbers to be utilized and rely on constraints of Quality of Services (QoS). Therefore, this paper has proposed a predictive VM placement using an efficient Learning Automata (LA) with probability distribution activity set and it can be represented as Probability Distribution Action-set Learning Automata (PDALA) which results to the VM placement over heterogeneous cloud DCs. Thus, the proposed algorithm gets beneficial by implementing LA theory and correlation coefficient parameter to generate best decision making over VM allocation. Moreover, CloudSim plus simulator is used to simulate results and the simulation output gets compared with Power Aware Best Fit Decreasing (PABFD) as reactive VM placement. The proposed PDALA method performance is evaluated using parameters like VM migration, SLA Violation and energy consumption having comparatively better performance than existing reactive VM placement.