High-Performance Computing

U Mass Dartmouth Cluster Features Hybrid Design for Diverse Workloads

• By Rhea Kelly
• 01/21/20

The University of Massachusetts Dartmouth has installed a new supercomputing cluster that will offer campus researchers five times the compute performance of the institution's previous setup. The cluster will support research in deep learning, astrophysical simulation, computational quantum chemistry, molecular dynamics simulation, solids analysis, computational fluid dynamics, systems security research and more.

The technology was deployed by Microway, a provider of computational clusters, servers and workstations for artificial intelligence and high-performance computing. Designed to serve a "wide array of campus workloads," the cluster features a hybrid architecture, according to a news announcement:

• More than 50 nodes include Intel Xeon Scalable Processors, DDR4 memory, SSDs and Mellanox ConnectX-5 EDR 100Gb InfiniBand;
• A subset of systems also feature NVIDIA V100 GPU Accelerators for GPU computing applications; and
• A second subset features POWER9 with 2nd Generation NVLink-based IBM Power Systems AC922 Compute nodes.

The advanced NVIDIA NVLink interfaces built into POWER9 CPU and NVIDIA GPU provide a broad pipeline between CPU:GPU for data-intensive workloads, the company said.

The cluster will be administered by the university's Center for Scientific Computing and Visualization Research.