Supercomputers Offered Through MCSR
MCSR offers access to two supercomputers.
Current Systems
Maple is a Cray cluster with a theoretical peak performance of 382 TFLOPS, 140 compute nodes, a total of 5,454 CPU cores, 29 NVIDIA Kepler K20 GPUs, 4 NVIDIA Tesla P100 GPUs, 12 NVIDIA Tesla V100S GPUs, and 27.1 TB of memory. This cluster was grant funded in part by the National Science Foundation (CHE-1338056).</>
Magnolia is an HPE Apollo 2000 cluster with 60 total compute nodes: 54 CPU-only nodes and 6 nodes with dual GPUs. The nodes are interconnected with HDR100 Infiniband. It runs Ubuntu Linux, uses the Bright Cluster Management software, and uses SLURM for job management.
All systems run SUSE Linux Enterprise Server and are connected to MCSR’s storage system.
Retired Systems
Sequoia was an SGI cluster. It consisted of 124 nodes with a total of 1304 CPU cores interconnected with a mixture of DDR and QDR Infiniband. Nodes had up to 35GB of memory and a varying number of CPU cores.
Sequoia will be phased out in the summer of 2020. As a general rule, new users should use Maple.
Sequoia is an SGI Altix XE cluster with 124 nodes, 1304 CPU cores, and 3,608 GB of memory with total performance of 13.68 TFLOPS. Each node contains two Intel Xeon processors. Nodes are connected using a mixture of DDR and QDR Infiniband. Parts of this cluster were purchased with funds provided by the National Science Foundation (EPS-0903787).
Node Details
- n001 – n042: 8 cores, 16GB
- n043 – n046: 8 cores, 31GB
- n047 – n124: 12 cores, 35GB
Catalpa was MCSR’s most recent single-image, shared memory system. It was an SGI UV 2000 with 320 processor cores and 2.5 terabytes of shared memory.
Catalpa is an SGI UV 2000 with 320 processor cores and 2.5 terabytes of shared memory. Unlike other MCSR systems, Catalpa is a shared memory system, not a cluster. It has a total theoretical peak performance of roughly 6.65 TFLOPS. Catalpa should be reserved for jobs that will not fit in the memory of the other systems.