Discovery  is a Linux cluster comprised of:

    • 85 16-Core (2x) Intel nodes (1360 cores)
    • 18 24-Core (2x) Intel nodes (432 cores)
    • 13 40-core (2x) Intel nodes (520 cores)
    • 10 64-core (2x) AMD nodes (640 cores)

In aggregate, the cluster has 2016 cores, 14.5TB of memory, and more than 1.8 PB of disk space.

Node Hardware Breakdown

Cell

Vendor

CPU

Cores

RAM

Disk

Scratch

Features

Nodes

g HPE Intel Xeon E5-2640V3 (2.6GHz) K80 X24 16
4992/card
128Gb
24Gb
1TB 820Gb gpu g01-g11
q HPE AMD EPYC 7532 (2.4ghz) 64 512Gb 1TB 820Gb amd q01-q10
j Dell Intel Xeon E5-2690  (2.6GHz) 24 96Gb 1TB 820Gb intel, cellj, ib3 j01-j18
k HPE Intel Xeon E5-2640V3 (2.6GHz) 16 64Gb 1TB 820Gb intel, cellk k01-k59
m HPE Intel Xeon E5-2643V4 (3.2GHz) 16 128Gb 1TB 820Gb test m01-m20
n HPE Intel Xeon Gold 6148 (2.40GHz) 40 384Gb 1TB 820Gb intel, celln n01-n13

Specialized Compute Nodes:

  • Discovery offers researchers the ability to have specialized head nodes available inside the cluster for dedicated computing. These nodes can come equipped with up to 64 compute cores and 1.5TB of memory

Operating System :

  • CentOS 7 is used on Discovery, its supporting head nodes and compute nodes.

Node Names :

    • The compute nodes for queued jobs will be managed via the scheduler.
    • GPU compute nodes are only available to members via the gpuq queue.
  • Interactive nodes are named x01 and are available for testing your programs interactively prior to submitting them to the queue to be run on the main cluster.

Node Interconnects

    • All of the compute nodes are connected via 10GB Ethernet. The cluster itself is connected to Dartmouth’s Science DMZ; facilitating faster data transfer and stronger security
  • cell j has 40GB Infiniband connections for parallel computing.