Evaluation of performance, energy consumption and cost for environments based on containers and virtual machines

Abstract

Virtualization has provided significant improvements in data center resource utilization, since reduces energy waste due to the reduced underutilization of physical machines in operation. However, the hypervisor factor and the virtual machine (VM) add unnecessary overhead to the overall performance of the host operating system, because each VM-based instance has its own guest operating system. On the other hand, containers have become a lightweight and flexible alternative to environments whose mandatory characteristics are agility and consistency in migration processes and application replication. Unlike traditional virtualization or paravirtualization technologies, containers do not require an emulation layer or hypervisor to run, instead they use an operating system-level interface. This feature makes them a lean technology as they require limited overhead, allowing higher instance densities to run on the host computer. This paper aimed to analyze CPU utilization and energy consumption costs in three experimental environments based on Docker containers and Kernel-based Virtual Machine (KVM). The results showed that the containers had a better balance of approximately 1.82% in CPU utilization, compared to 28.89% of VMs. Besides, the energy consumption and cost per killowatt hour of containers were significantly lower.