Version Control Graphical Interface for Open OnDemand

Abstract

Use of high performance computing (HPC) clusters is challenging for the average researcher who is not familiar with the necessary tools such as the Linux operating system and job management software. Of those using HPC systems, the majority are not trained specifically in computer programming. Nearly 70% of users on the NSF-funded XSEDE system are in fields other than computer science and computing science. The graphical user interfaces (GUIs) that most users rely on for common software on personal computers are typically not available on HPCs. The Ohio Supercomputer Center has addressed these issues through the development of an open source, web-based GUI called Open OnDemand. Because it is open source, administrators are free to deploy it on their own systems, and developers are free to enhance it. To improve workforce development and diversity, we sought to: (1) install Open OnDemand on a private cluster and (2) develop a custom add-on module for version control of HPC application software. We successfully installed and configured Open OnDemand for a private cluster at Case Western Reserve University in order to evaluate the difficulty level of deployment. Despite our lack of experience in HPC system administration, the installation process was straight forward due to a streamlined installer package and thorough documentation. In order to evaluate the extensibility of Open OnDemand, we developed a custom web-based interface for use of the popular Git version control system. Version control tools are important in maintaining software by easily tracking and accessing file changes in both single- and multi-developer projects. Our module successfully integrates with the existing Open OnDemand interface and provides common version control operations that can be used during typical HPC workflows. It is our hope that making version control software easier to use will encourage HPC users to adopt the use of version control into their own workflows to improve productivity and repeatability. Source code for the app will be made available on the author's github site.