MCC system overview

MCC system overview

The Morgan Compute Cluster (MCC) is a high-performance computing system designed for large-scale scientific workloads.
The system provides high-core-count AMD compute nodes connected through high-speed networking and shared parallel storage.



Admin Nodes (Motherships)

Administrative nodes manage cluster services such as scheduling, monitoring, and system management.

Processor

Processor Class

Cores per Node

Nodes in Cluster

Memory per Node (GB)

Network

Node Names

Processor

Processor Class

Cores per Node

Nodes in Cluster

Memory per Node (GB)

Network

Node Names

Intel Xeon Gold 6230

Cascade Lake

64

2

92

InfiniBand HDR (100 Gbps)

mothership2[1-2]

Login Nodes

Login nodes provide access points for users to connect to the cluster, compile software, and submit jobs.

Processor

Processor Class

Cores per Node

Nodes in Cluster

Memory per Node (GB)

Network

Node Names

Processor

Processor Class

Cores per Node

Nodes in Cluster

Memory per Node (GB)

Network

Node Names

AMD EPYC 7702P

Rome

64

2

256

InfiniBand HDR (100 Gbps)

mcc-login[001-002

Compute Nodes

Compute nodes perform the primary scientific workloads on the cluster.

Node Type

Processor

Processor Class

Cores per Node

Nodes in Cluster

Total Cores

Memory per Node (GB)

Network

Node Names

Node Type

Processor

Processor Class

Cores per Node

Nodes in Cluster

Total Cores

Memory per Node (GB)

Network

Node Names

AMD Compute

EPYC 7702P

Rome

128

180

23,040

512

InfiniBand HDR (100 Gbps)

rome[001-180]

Large-memory Compute

EPYC 7702P

Rome

128

2

256

4096

InfiniBand HDR (100 Gbps)

frome[001-002]

Data Transfer Node (DTN)

Data transfer nodes are used for efficient movement of data into and out of the cluster.

Processor

Processor Class

Cores per Node

Nodes in Cluster

Memory per Node (GB)

Network

Node Name

Processor

Processor Class

Cores per Node

Nodes in Cluster

Memory per Node (GB)

Network

Node Name

AMD EPYC 7702P

Rome

64

1

256

Ethernet (10 Gbps external), HDR (100 Gbps internal)

mcc-dtn

 

Storage

The MCC cluster uses a shared parallel filesystem for high-performance data access across compute nodes.

  • Lenovo GPFS (DSS-G) parallel filesystem

  • 1.8 PB usable capacity (2.3 PB raw)

  • Connected via 200 Gbps HDR fabric

Related Information

For details on SLURM queue names see the SLURM Queues page.

 

Center for Computational Sciences