What is the largest HPC cluster you've run? RRS feed

  • Question

  • Am interested in how large others have scaled up HPC clusters. In terms of both total computenodes, and total cores.

    Is 1,000 nodes feasible? Or 10,000 cores (across several hundred nodes)?

    Was this with a single headnode? Or clustered?

    Thanks in advance!

    Tuesday, July 26, 2016 4:27 PM


  • Hi TimJRoberts1,

    We officially support 1,000 nodes and 16,000 cores (16 cores per node X 1000). In our scale test environment, the nodes and cores sometimes may reach 2,000 nodes and 20,000 cores. It is feasible to have a single head node instead of clustered high-availability head nodes for easy setup and maintenance, however a remote SQL database with scale configurations and high hardware spec machines (e.g. 2*E5-2660V2 @ 2.20GHz, 12*16=192GB RAM, 8* 600GB 15K RPM HDD, 1Gb NIC) are required for the large scale cluster.


    Yutong Sun

    • Marked as answer by TimJRoberts1 Wednesday, July 27, 2016 8:02 AM
    Wednesday, July 27, 2016 1:54 AM