Towards Docker Live migration

Abstract

The objective of this thesis was to find out how we can do live migration of a running process using a lightweight container management system known as “Docker”. We will introduce the concept of virtualization with details structure as well as the use of Docker. Live migration is experimental feature of Docker. There are some certain conditions to set up the system environment; our aim is to discuss it in details. We have used 64 bit UBUNTU 16.04.4 LTS as the host OS and docker version 17.09.1~ce is used with experimental feature enabled. We have showed how we have to prepare source and destination environment and all the steps to successfully execute live migration. This live migration requires a time which is called as migration time, transfer time or delay time. We have also calculated this transfer time for different type of source and destination. After that, we have compared and analyzed the data which shows that for a same migration process that transfer time can be different. Also, a graph based on the experimental data shows that the tarball size used for the migration increases proportionately with the time.