Servers and storage

STORAGE

  • Storage for production on DELL ISILON, ~650 Tb usable composed of 6 H400 nodes
  • Storage based on IBM Spectrum Scale for scratch (ex GPFS), ~300TB usable with 2 IO nodes [Dell R730] + JBOD [Dell MD3460]
  • Storage for production storage replication, GlusterFS, remote site, 400TB
  • Storage based on ZFS for backup, 50 TB (mirrored on remote site)

COMPUTING

  • Cluster cortex
    Total of 25 compute nodes, 512 cores, 7.2TB RAM
    24 standard compute nodes, Dell R630 (20 cores, 256 GB RAM)
    1 "big memory" compute node, Dell R630 (32 cores, 1TB RAM)
    2 login nodes, Dell R630 (20 cores, 256 GB RAM)
  • GPU server
    2 x Intel Xeon Silver 4114 (10-core)
    128 GB RAM, SSD disks (3To)
    8 x NVIDIA TESLA T4 16GB GDDR6

VIRTUALIZATION

  • Cluster Proxmox VE
    3 servers DELL PowerEdge R640 (2 CPU Xeon 4114, 20 cores, 256GB) + dedicated shared storage (20TB)

NETWORK

  • Internet access 1 Gbits/s (shared), University of Bordeaux
  • LAN Intel Omni-Path (data), 100Gbit/s
  • LAN Ethernet (administration), 10 Gbits/s

EXTERNAL COMPUTING RESOURCES

HOSTING INFRASTRUCTURE

  • IT Room (28m²)
  • Hot aisle containment
  • Chiller + FreeCooling (~50% of the time)
  • Heat exchangers air/water (water at 15°C / 20°C)
  • UPS fault tolerance capacity of 48 KWatt N+1
  • Supervision of the IT room (temperature, humidity, water leakage, consumption by outlet, UPS, chiller, LCP)