HPC cluster Zhores

 

 

  • General description
    Vendor: DELL/Supermicro
    Platform: Hybrid cluster platform
    General purpose: for research requiring a large amount of hybrid computing resources: GPU and CPU
  • The compute part
    Compute blades:
    CPU Nodes – 48
    MEM nodes – 10
    GPU Nodes – 26
    Virtualization nodes – 6
    Peak performance 1 PFLOP/s
    104 GPUs(Tesla V100-SXM2) with 16Gb of memory
  • Interconnect
    Data network: InfiniBand EDR (100 Gb/s)
    Service and management networks: 1GbE
  • The data storage system
    Parallel file system of data storage: GPFS 416T
    Home directory: BeeGFS 100T
    CephFS
  • OS
    CentOS Linux release 7.9 (kernel 5.4.15)
  • Workload Manager
    Slurm 20
  • The list of available queues
  • The list of software
    • Publication :

‘Zhores’ — Petaflops supercomputer for data-driven modeling, machine learning and artificial intelligence installed in Skolkovo Institute of Science and Technology

If you have a technical questions regarding the cluster, please send your question(s)/request(s) to: