The thesis introduces the development of an HPC system which uses Docker. It introduces
the requirements, the applied technologies, the created architecture, the used technical
solutions and the implementation as well.
The goal of the whole development process was to upgrade Superman, which is the
university’s supercomputer used for academic research as well as resolving existing oper-
ational problems, and the prevention of possible future problems.
The document also analyzes the solutions used by the current system and introduces
new requirements based on the gained experience. It also describes the selected tech-
nologies used for the implementation: the structure and use of the HTCondor system, the
container technology — linux namespaces, cgroups, etc. It introduces the Docker container
and how to use it. In addition, it also introduces the VXLAN network protocol which is
used for establishing network between containers. It describes the Ansible configuration
management tool, which helps to install dependencies and set configuration files.
It discusses the current system’s operational problems and the used workarounds for
solving them. It also describes the implementation details of the new system. It presents
the environment used for development and testing and shows an implementation example
which can be used for making future expansion to the system.