The University of Arkansas for Medical Sciences maintains high performance computing facilities for research use by faculty and staff. The Information Technology department manages the data centers and networks used to access these facilities, and provides low level administrative support. The IT department also maintains a private cloud (VM infrastructure) that is used by the high performance computing facilities in various support functions. The Department of Biomedical Informatics handles the administration of the high performance compute facilities themselves.
Our primary system is named Grace, after Rear Admiral Grace Hopper, a pioneer in computer science. Grace is composed of 96 traditional Xeon CPU nodes each with 28 cores and 128 GB of memory (2688 CPU cores), 96 Xeon Phi nodes each with 64 cores, of which 80 have 384 GB of memory and 16 have 192 GB of memory (6144 Phi cores total), 3 Xeon nodes with 24 cores and 128 GB of memory plus 2 NVIDIA GPUs each (72 CPU cores, 6 GPUs with 27,264 total cores), 7 management/login/storage interface nodes and 1.9 PB of high speed storage (DDN GS14KX GridScaler). All components are interconnected via 100Gbps Omnipath interconnects and are attached to external storage and the Internet (both commercial and I2) with redundant 10Gbps Ethernet interfaces. Intermediate storage is provided by a 1PB NAS and long term/backup storage is provided by a 4.2 PB object storage system (Dell/EMC ECS appliance).
All high performance computing resources are made freely available to UAMS researchers. Advanced open source software packages (including packages developed by the UAMS DBMI) to support genomic, metagenomic, proteomic, metabolomics, microbiomic and image analyses are installed on the system. Additional packages can be installed or investigators can run pipelines implemented in Singularity containers.
UAMS also has direct connections from the UAMS campus network to the High Performance Computing facilities hosted at the University of Arkansas in Fayetteville. Instructions for accessing those systems from UAMS can be found here. The University of Arkansas in Fayetteville and the University of Arkansas for Medical Sciences have signed a Memorandum of Understanding (MOU) forming the Arkansas Research Computing Collaborative (ARCC) in order to share, enhance and manage existing and future high performance and cloud computing resources and associated mass storage capabilities.
Space contributors
{"mode":"list","scope":"descendants","limit":"5","showLastTime":"true","order":"update","contextEntityId":57311274}