Resource Allocation Algorithms for Single, Cluster and Tired Virtual Machines
Andrew V. Toutov, Natalia V. Toutova, Georgy Bulanov, E. A. Frolova, I. A. Andreev · 2023
Data centers must provide sufficient resources for the smooth operation of applications hosted in them in conditions of variable workload. Resource management is a necessity for guarantee SLA (Service Level Agreements). This paper provides an overview of methods and algorithms for virtual machine allocation in cloud data centers based on the analysis of recent work. The advantages and disadvantages of existing approaches are considered, as well as the requirements that must be met by a universal method of hosting virtual machines that allows overcoming existing limitations. The cloud data center resource management process consists of the following steps: server monitoring, identification of overloaded/underloaded servers, virtual machine selection for migration, destination host selection. Two algorithms for virtual machine selection for migration and destination host selection are proposed. They are suitable for single, cluster and tired virtual machines. These algorithms allow to find an exact solution for the combined criterion of energy efficiency and SLA violation, which is on average 3.5 times better than the widely used FFD (First Fit Decreasing) and BFD (Best Fit Decreasing) heuristics.