HPE Hunter Hardware and Architecture

With respect to some technical details of Hunter Hardware and Architecture, please refer to this Slides.

Summary

Node/Processor

hunter APU compute nodes

• Blade: HPE Cray EX255a (El Capitan blade architecture, MI-300A)
• APU: AMD Instinct MI300A Accelerator
• 4 APU's per node
• 24 CPU Cores and 228 Compute Cores per APU
• Memory: 512 GB per node (HBM3)
• HBM3 ~5.3 TB/s per APU
• Network: HPE Slingshot 11 (4 injection ports per node, 4x200Gbps)
• Number of nodes: 188
• some nodes have 1 local NVMe M.2 SSD installed.

Pre- and post processing

Within HLRS simulation environment special nodes for pre- and post processing tasks are available. Such nodes could be requested via the batch systems using the queue "pre" or queue "smp". Available nodes are

    4 nodes 3 TB Memory, 2 Socket AMD EPYC 9354 32-Core Processor, exclusive usage model, available by queue "pre"
    1 node  6 TB Memory, 2 Socket AMD EPYC 9354 32-Core Processor, shared usage model,    available by queue "smp"

more specialized nodes e.g. graphics, vector, DataAnalytics, ... are available in the Vulcan cluster.
If you need such specialized nodes on vulcan cluster for pre- or postprocessing inside your project located on hunter resources, please ask your project manager for access to vulcan.

Interconnect

HPE Slingshot 11 Dragonfly

Filesystem

Available lustre filesystems on hunter:

• ws12 (HPE Cray ClusterStor E1000 Lustre Appliance):
  • available storage capacity: ~12 PB
  • lustre devices: 2 MDT, 20 OST
  • performance:

Note: In a further installation phase this year, an HPE Cray ClusterStore E2000 Lustre appliance with a capacity of 13 PB will additional be installed. In a third installation phase, the current HPE ClusterStore E1000 will be upgraded with an E2000).

additional an central HOME and project fileserver is also mounted on hunter. Some special nodes have a local disk installed which can be uses as localscratch.

See also Storage (Hunter)