HPC project for Leading Academic Institute - Bangalore

Executive Summary:

The case study involves a leading Academic Institute in Bangalore rated among the Top 3 in the country. The Institute is involved in imparting world class higher education in an environment of fundamental and applied research in science and engineering disciplines that is globally reputed.

Customer Background:

The department involved in Molecular Modeling and Simulation had a few small HPC Clusters, CPU based for running the various MD Simulation packages. The cluster soon was saturated with various applications and was giving a slow response as the problems grew in size. A need was felt not only to set up a larger HPC system but also to use the price and performance benefits of GPU technology. The system size was conceived to meet the computing needs of the department for at least 5 years.

Solution Overview

The HPC Configuration was a Hybrid system ( involving both CPU and GPU Nodes) with large high speed Storage, a high bandwidth, low latency network to interconnect all the nodes.

The configuration consists of the following:

1. Master Node – Dual AMD EPYC 64 core processors
2. CPU based compute nodes – 13 – dual AMD EPYC 64 core CPUs per node
3. GPU Nodes – 5
4. High Speed Lustre Storage – 1.2 PB Storage capacity
5. 100 Gbps low latency, high bandwidth inter-node interconnect fabric
6. Software: CentOS Open Source Linux, AMD compilers, NVIDIA CUDA tools, various Molecular Dynamic Simulation packages such as Gromacs, NAMD, Amber, LAMMPS etc.