https://hpcc.ucr.edu/about/hardware/Recent content in Hardware on HPCCHugo -- gohugo.iohttps://hpcc.ucr.edu/about/hardware/overview/Mon, 01 Jan 0001 00:00:00 +0000https://hpcc.ucr.edu/about/hardware/overview/ <h2 id="computer-cluster">Computer cluster</h2> <ul> <li>Over 16,000 CPU cores (60% AMD, 40% Intel)</li> <li>512-1024GB RAM per node</li> <li>GPU: 24x K80, 2x P100, 24x A100, 2x H100</li> <li>IB network @ 56Gbs - 200Gbps (FDR, HDR and NDR)</li> <li>Queueing system Slurm</li> </ul> <h2 id="parallel-data-storage-system">Parallel data storage system</h2> <ul> <li>5PB GPFS storage (scales to &gt;50PB)</li> <li>Home directories on dedicated system</li> </ul> <h2 id="backup-system">Backup system</h2> <ul> <li>5PB GPFS storage</li> <li>Geographically separated server room</li> </ul> <h2 id="server-room">Server room</h2> <ul> <li>School of Medicine Server Room</li> <li>Size 600+ sqft</li> <li>Raised floor cooling with redundant AC units</li> <li>Backup power: UPS plus generator</li> </ul>https://hpcc.ucr.edu/about/hardware/details/Mon, 01 Jan 0001 00:00:00 +0000https://hpcc.ucr.edu/about/hardware/details/ <h2 id="storage">Storage</h2> <ul> <li>Four enterprise class HPC storage systems</li> <li>Approximately 3 PB production and 3 PB backup storage (total 6 PB or 6,144 TB)</li> <li>GPFS (NFS and SAMBA via GPFS)</li> <li>Automatic snapshots and monthly backups</li> </ul> <h2 id="network">Network</h2> <ul> <li>Ethernet <ul> <li>2 x 40 Gb/s redundant connections to the campus high performance research network</li> <li>1 x 10 Gb/s connection to the public internet</li> <li>Redundant, load balanced, robust mesh topology</li> </ul> </li> <li>Interconnect <ul> <li>56/200/400 Gb/s InfiniBand (FDR/HDR/NDR)</li> </ul> </li> </ul> <h2 id="head-nodes">Head Nodes</h2> <p>All users should access the cluster via SSH through <code>cluster.hpcc.ucr.edu</code>. This address will automatically balance traffic to one of the available head nodes.</p> <ul> <li>bluejay, skylark <ul> <li>Resources: 256 cores, 512 GB memory</li> <li>Primary function: submitting jobs to the queuing system</li> <li>Secondary function: development; code editing and running small (under 50 % CPU and under 0.5 GB RAM) sample jobs</li> </ul> </li> </ul> <h2 id="worker-nodes">Worker Nodes</h2> <p>For a full list of nodes, please see our <a href="https://docs.google.com/spreadsheets/d/1SVH1-c1i075vjt-B0wNPiK87wmLkPltWJlIPgLkmoqU/">Node List</a> sheet. This contains info on all nodes (public and private) that we have. This includes CPU, memory, and GPU specs, as well as any features that the nodes have.</p> <p>For more info about queueing jobs see the <a href="https://hpcc.ucr.edu/manuals/hpc_cluster/jobs/">Managing Jobs</a> page. For queueing jobs to nodes with specific features, see the <a href="https://hpcc.ucr.edu/manuals/hpc_cluster/jobs/#feature-constraints">Feature Constraints</a> section.</p>