Performance effects on servers using containers in comparison to hypervisor based virtualisation

Abstract

The current direction of web-based software is evolving has brought many challenges regarding locations, scalability, maintainability and performance. In most of the cases tiny differences in performance are not really important since the user base of the software may not be remarkably high. However, when the user base of the system expand and the stress of the system has high peaks, smaller things start to have meaning in the software and infrastructure. This paper addresses the performance differences between todays usual web software deployment solutions, containers and VMs.

This study is based on literature review, which has been conducted by studying previous studies about the topic. The main focus in the study is Linux Container based containerization solutions such as Docker, and traditional hypervisor based virtual machines such as KVMs.

The main categories of performance this paper is addressing to are memory, CPU, network, disk I/O, database applications, DevOps and overall system performance. The categorizing is based on the studies reviewed in this paper which obviously results in slight overlapping on some categories. In the end of this paper the results are summed up and some implications for them are presented.