Cluster Overview¶
Research CycleCloud (RCC) is a shared high-performance computing environment used for research computing workloads.
Basic Architecture¶
RCC includes:
- A head node, also called the scheduler node
- A shared filesystem
- Multiple compute nodes across partitions
- Posit Workbench access for RStudio, Jupyter, and VS Code
- A limit of 1800 active cores in parallel
Common Slurm partitions include:
ccdefaulthighmemdragen_v3dragen_v4gpu
Basic Components¶
Node¶
A node is a virtual machine within the cluster. It contains processors, also called CPUs, and memory, also called RAM. Some nodes may also have GPUs.
Core¶
In the RCC context, a CPU is treated as a core.
ccdefault Partition¶
The ccdefault partition includes compute nodes with 4 to 48 CPUs and 4 GB RAM per CPU.
| Slurm Partition | SKU | CPUs | Memory | Scratch Space | Cost per Hour |
|---|---|---|---|---|---|
| ccdefault | d4 | 4 | 16 GB | 32 GB | |
| ccdefault | d8 | 8 | 32 GB | 64 GB | |
| ccdefault | d16 | 16 | 64 GB | 128 GB | |
| ccdefault | d32 | 32 | 128 GB | 256 GB | |
| ccdefault | d48 | 48 | 192 GB | 384 GB |
Specialized Partitions¶
Specialized partitions provide access to larger memory, DRAGEN, and GPU resources.
highmemhas 3 nodes with 48 CPUs and 16 GB RAM per CPUdragen_v3anddragen_v4allow access to Illumina DRAGENgpuhas an NVIDIA Tesla T4 GPU processor- These partitions are subject to the same overall usage limit as
ccdefault
| Slurm Partition | SKU | CPUs | Memory | Scratch Space | Cost per Hour |
|---|---|---|---|---|---|
| highmem | M64 | 64 | 1 TB | 2 TB | |
| dragen_v3 | NP20 | 20 | 336 GB | 1.47 TB | |
| dragen_v4 | NP20 | 20 | 336 GB | 1.47 TB | |
| gpu | NC16 | 16 | 110 GB | 352 GB |