Dynamic Resource Provisioning in Datacenters using Profitability-aware VM Placement

With the ever-spreading user-base in cloud datacenters, it has become increasingly challenging to satisfy each user’s service demand. Resource requirements or service demands of cloud users are embedded in virtual machines (VM). Virtual machines are software extensions of physical machines such that a single physical machine can generate a number of virtual machines and hence, can service multiple users. VM placement is the task of allocating a VM to a chosen physical machine (PM) such that the resource demands of that VM are satisfied by the physical machine. An effective VM placement aims at balancing load in the datacenter thereby avoids unnecessary VM migrations and improves the performance of a datacenter. Bin packing has been widely used as a technique for VM placement in the variations of first-fit, best-fit and/or worst-fit. This paper proposes a novel VM placement technique which ensures that the profits to the cloud service provider are maximized, establishes fairness and service availability to the cloud user by dynamically placing a virtual machine to the best available physical machine using the concept of clusters. Our VM placement strategy avoids unnecessary migrations in the datacenter by balancing load on all the physical machines present inside a cluster. A comparison, of the proposed technique with First-Fit bin packing placement technique, highlights a better time efficiency and resources utilization in the present work.