Henson High Performance Computing Lab
The Henson High Performance Computing Lab is a Beowulf cluster designed for both classroom instruction as well as a research cluster. During the day, it is used in the instruction of high performance computing, scientific computing, computer graphics, and general Linux instruction. In the off-hours, the cluster software is activated to research jobs to run either overnight or during the weekend. The Henson HPCL furthermore acts as the administrative unit for all Linux support and Linux resources for the Henson School.
About the HPCL
The HPCL cluster is composed of 34 high-performance computer nodes. Each node features a 512GB NVMe SSD, 32GB of DDR5 4800MHz RAM, and a shared /home directory.
All 34 nodes are equipped with an NVIDIA GeForce RTX 4080 (9,728 CUDA Cores) and a 13th Gen Intel(R) Core(TM) i9-13900KF CPU (24 Cores / 32 Threads), making them all capable of CUDA computation. The total computational capability of the cluster consists of 816 physical cores / 1088 threads, a total scratch capacity of approximately 17.4TB, and just over 1TB (1088 GB) of DDR5 RAM.
The HPCL cluster is primarily designed to run MPI-aware applications and also supports running and deploying large language models with Ollama; however, the cluster management software remains capable of supporting applications with alternative considerations such as applications requiring license servers, alternative network configurations, or batch processing applications. Additionally, there is experimental Hadoop support using the Magpie framework for deployment of Hadoop via the cluster management software.
Beyond its core MPI capabilities, the HPCL cluster also provides support for running and deploying large language models with Ollama, as well as robust computation and statistical analysis using R and the R Studio environment.
Access to the Henson HPCL cluster requires authorization. To request access, please contact the IT System Administrator for the Henson School, Jungho An. You can reach him at jan@salisbury.edu.
