Cluster Resources
FEDGEN Datacenter HPC resources includes Compute, GPUs and Storage Resources which are being increased on an annual basis. The FEDGEN HPC cluster has very similar hardware with specific processors in variety of configuration in memory capacity or number of CPUs. All the nodes share resources over networks for data storage and cluster communication.
Resource Manager
FEDGEN HPC Uses SLURM Resource and Queue Manager Software to manage Cluster Resources. Jobs submitted to SLURM job queue can run on any possible nodes dependent on job or cluster constraints. Users only have to specify resource requirements and SLURM scheduler will assign the job to an appropriate node or set of nodes.
Operating Software
FEDGEN HPC Cluster currently runs on Ubuntu 22.04 .
Node Info (Compute/GPU)
Cluster Type |
Node Count |
Processors (Architecture) |
Cores |
Memory |
Disk Size |
GPUs (Number) |
Node Name |
Model |
|---|---|---|---|---|---|---|---|---|
Fedgendc1 |
3 |
Intel Xeon Gold 5218R Hyperthreading Enabled (Cascade Lake SP) |
80 |
512 GB |
900 GB |
giga-[001-003] |
SuperMicro |
|
3 |
Intel Xeon Bronze 3408U (Sapphire Rapids) |
8 |
16 GB |
500 GB |
Nvidia A2000 12 GB memory (1) |
terx-[004-006] |
SuperMicro |
|
1 |
Intel Silver 4210R (Cascade Lake) |
20 |
512 GB |
1000 GB |
Nvidia A100 SXM with 40 GB Memory (1) |
terv-[007] |
ASUS |
|
1 |
Intel Xeon Platinum 8358 (Cascade Lake) |
128 |
1024 GB |
1.92 TB |
Nvidia A100 SXM with 81920 MB memory (3) |
tera-[008] |
SuperMicro |
Networks (Interconnect)
All nodes are connected to a 10Gb Ethernet network for Cluster Communications as well as a 2x 10Gbps Ethernet network for storage access
Storage
Local Scratch space ~ High Speed SSD Drive in RAID 1 (Available on each Node) Size is dependent on node
User Home Directory ~ 4TB on network file system (NFS) available on the login node, the compute nodes and the GPU nodes.
Shared Scratch space ~ 14 TB network file system (NFS) provisioned from storage node
For more details see Storage section.