Slurm "features" are available on every node to allow users to request specific types of hardware.  


Details for the UB-HPC cluster include:


CPUs:

INTEL

AMD

and also specific CPU model numbers such as: 

CPU-Gold-6230

CPU-Gold-6130

CPU-6132HE

CPU-L5630

CPU-L5520

CPU-E7-4830

CPU-E5-2660

CPU-E7-4830

CPU-X7550

CPU-E5645

CPU-Phi-7250



GPUs:

V100

P100

gpu:tesla_v100-pcie-16gb:2(S:0-1)

gpu:tesla_v100-pcie-32gb:2(S:0-1)

gpu:tesla_p100-pcie-16gb:2(S:1)

Available on the visualization nodes through OnDemand interactive desktop sessions only:

gpu:quadro_p4000:2(S:0-1)

gpu:quadro_k4200:3(S:0-1)

gpu:quadro_p4000:4



To select a specific high-speed interconnect network:

IB (Infiniband)

OPA (OmniPath)



To select nodes only from certain funding sources (recent grant purchases):

NIH - formerly cascade partition

MRI - formerly skylake partition



Machine room location - to request nodes in certain racks, examples include:

u22

u25

d07

k10



Details for the Industry partition within the UB-HPC cluster include:


CPUs:

CPU-Gold-6330


GPUs:

A100

gpu:a100-pcie-40gb


To request a node with a high-speed interconnect network:

IB (Infiniband)


Machine room location - to request nodes in certain racks, the industry cluster is in racks:

h22

h23

h24

h25



Details for the Faculty cluster:


The faculty cluster is made up of more than 50 partitions that include nodes purchased as recently as this year and as long ago as 8 years ago.  It's a heterogenous mix of architectures and features.  The best way to get information on the faculty cluster is using the snodes command.  See below.




The best way to see all the Slurm features offered and what is currently available, is by running the 'snodes' command in the terminal on one of the cluster front end servers.  More details


The format of the Slurm features in snodes command output is:

CLUSTER, CPU_MODEL, CPU_MANUFACTURER, RACK,[FUNDING SOURCE, INTERCONNECT, GPU_MODEL] 

Anything in [ ] is optional and may be dependent on what hardware is in the node




How to request specific hardware for your jobs