Distributed Algorithm for Balanced VM Placement for Heterogeneous Cloud Data Centers

Virtual Machine (VM) placement aims to efficiently deploy the dynamically arriving VM requests on active physical servers of a data center. It is one of the most challenging and NP-complete problem [9] while considering the large number of optimization criteria and multi-resource VM requirements. In this context, most of the existing works deal only with limited trade-off among types of resources, individual criteria of optimization and homogeneous computing environment, thus resulting unnecessary activation of servers and drastically increased energy consumptions. High energy consumption will not only lead to high operational cost but also maximize the CO2 emissions to the surroundings. In order to achieve the objectives of energy-efficiency, cloud is required to be analyzed with the perspective of balanced resource utilization. Thus, in this work we propose a distributed multi-coloring model for balanced VM placement and goes beyond the current state of the art by maximizing load balancing and efficient use of computing resources. The proposed solution is extensively evaluated through simulation. Experimental results demonstrate the effectiveness of solution over the existing heuristics and show that it can minimize the number of running servers effectively while optimizing the performance metrics of energy saving, resource utilization and load balancing.