Containers is a tool which is often used in development and testing applications because they are easyto-use and lightweight. The container is built based on an image, which is a template for the container
and a way to transmit it over the network. Images can be up to several gigabytes in size. Thus, if you
need to transfer the large application container over the network, or the available memory is scarce
(for example, when developing IoT systems), it is necessary to minimize the size of the image.
Container images are built on top of the Linux kernel. For interactive work with containers, they
include instructions and files that the application never works with inside the container. The
applications that will be launched in the container and their number are known in advance, so you can
use Linux debugging tools to find out which files are used by applications in the container, and which
can be excluded from it. This is how the Chainsaw application works. This paper presents the results
of a study of several different Docker container images. For some application we obtained threefold
decrease of the image size; also, we found that the size of the reduced image is significantly affected
by the base image and the programming language in which the application that runs inside the
container is written.