AHPSelection: aplicação da técnica AHP no problema de máquinas virtuais para migração em um ambiente de data center

Abstract

Cloud Computing is a new computational model that has revolutionized the area of Information Technology (IT). Such a paradigm allows the on-demand supply and consumption of computational resources present in large infrastructures known as data centers in the form of services. The success of this model has led to an increase in the number of data centers, which has led to the energy consumption of such infrastructures reaching worrying levels. One way to minimize this power consumption is by intelligently distributing Virtual Machines (VMs) among its component servers. Such distribution comprises the use of Allocation Policies in choosing the best server to receive a VM and a Selection Policy in choosing a VM for allocation or migration. This paper presents a new VM Selection Policy, AHPSelection, which maps the process of choosing a VM for migration as a decision making problem and applies the Analytic Hierarchy Process (AHP) technique as a tool to assist in this procedure. Initially, AHPSelection presented intermediate results in terms of performance and energy consumption. This behavior led to the creation of a modified version of AHPSelection (called AHPSelection2) that showed good results in terms of power consumption and number of migrations in scenarios where there is little variation in VM resource demand. AHPSelection2 was able to reduce by four to five times the total number of migrations in some scenarios, as well as providing an average energy savings of 15 Kwh in others.