Skip to main content

Continuing the good work at Brandeis University

A team from the IU Pervasive Technology Institute - UITS Research Technologies assisted in upgrading an existing HPC cluster at Brandeis University in Waltham, MA.

Extreme Science and Engineering Discovery Environment (XSEDE) Capabilities and Resource Integration staff from Indiana University assisted in upgrading an existing HPC cluster at Brandeis University in Waltham, MA. Francesco Pontiggia, who was working with the XSEDE Capabilities and Resource Integration group as and XSEDE Champion Fellow, was managing a cluster of 145 nodes, running on Rocks 6. Given the lack of forthcoming updates, he wanted to move to a more modern operating system, and asked for help from XCRI in updating his cluster, while also providing a friendly-user implementation opportunity to the XCRI team. Over the course of a week, Francesco worked very hard with XCRI Engineer Eric Coulter, IU, to implement the new OpenHPC-based version of the XSEDE Compatible Basic Cluster, resulting in great improvements to both the XCBC project and the management of the Brandeis cluster. The Brandeis system is in flux, but when completely migrated to OpenHPC, it will have 145 nodes with 1900 cores, and 168 GPUs. Jason Wells from Bentley University, who collaborated with XCRI the previous year in building a new HPC system, visited for a few days during the build as well, highlighting the ways that pushing for standards in HPC can engender community among cyberinfrastructure professionals.

 HPC Professionals at Brandeis University

Francesco Pontiggia, Brandeis Research Computing Specialist, and Jason Wells, Bentley University Senior Instructional and Research Technology Manager. Photo credit Eric Coulter, XCRI Engineer

As a part of its Campus Bridging effort, XSEDE distributes a custom Rocks roll designed to minimize the complexity of building XSEDE-compatible Linux clusters for use by the US open science community. The XSEDE Rocks roll provides the widely used, open-source, scientific, mathematical, and visualization packages needed to convert a "bare-bones" Rocks cluster into an XCBC, a high-performance parallel computing cluster that's compatible with XSEDE digital services.

This was a very useful build for both XCRI and the Brandeis Scientific HPC center. As a result of this trip, XCRI developed a much more seasoned toolkit, and gained valuable experience with in-place migration of hardware from a running system to an updated version. This was also a necessary step for Brandeis, as the operating system in use on the old system was rapidly approaching end-of-life, and would no longer receive the updates needed for current versions of scientific software.