QCL Computational Resources
High Performance Computing (HPC)
Supercomputing Allocations through ACCESS (formerly XSEDE)
ACCESS is a collection of integrated advanced digital resources and services that provide easy access to the most advanced computational resources and scientific research support in the world. QCL has two Campus Champions who are trained to help local users utilize supercomputing resources available through the Access program. From various national supercomputing facilities, QCL has awarded a large number of computing hours for testing and developing scientific applications (see below).
ACCESS Campus Champion Allocations (As of January 2023)
| Name | Facility | SU (Core hours) |
|---|---|---|
| Rockfish - GPU | Johns Hopkins | 500 GPU hours |
| Rockfish - Large Memory | Johns Hopkins | 1,000 Core hours |
| Rockfish - Regular Memory | Johns Hopkins | 20,000 Core hours |
| KyRIC Large Memory Nodes | Kentucky Research Informatics Cloud | 1,000 Core hours |
| Jetstream | Indiana U | 50000 SUs |
| Bridges-2 Regular Memory | PSC | 50,000 SUs |
| Bridges Extreme Memory | PSC | 1,000 Core Hours |
| Bridges-2 GPU | PSC | 2,500 GPU Hours |
| ANVIL CPU | Purdue | 100,000 SUs |
| ANVIL GPU | Purdue | 1,000 SUs |
| Stampede2 | TACC | 1,600 Node Hours |
| EXPANSE CPU | SDSC | 50,000 Core Hours |
| EXPANSE GPU | SDSC | 2,500 GPU Hours |
| DARWIN Compute Node | UD | 20,000 SUs |
| DARWIN GPU | UD | 400 SUs |
| OSG | Multiple | 200,000 SUs |
The Service Unit (SU) is like a currency used to run an application on supercomputers. Supercomputer users are charged by SUs (hours of runtime on one core). For example, if an application ran on 100 cores for 10 hours, 1,000 SUs will be deducted from your account.
The Campus Champion allocations can be used to test computational research applications. To test out the supercomputers, please make an appointment with one of the CMC Campus Champions (email: qcl@cmc.edu).
NVIDIA GPGPU Machine
A GPGPU (General Purpose Graphic Processing Unit) machine is a high performance computer system equipped with one or more GPUs. QCL has an NVIDIA DGX system having four Tesla V100 GPUs.
| Component | Spec |
|---|---|
| GPUs | 4X Tesla V100 |
| TFLOPS (Mixed precision) | 500 |
| GPU Memory | 128 GB total system |
| NVIDIA Tensor Cores | 2,560 |
| NVIDIA CUDA Cores | 20,480 |
| CPU | Intel Xeon E5-2698 v4 2.2 GHz (20-Core) |
| System Memory | 256 GB RDIMM DDR4 |
| Data Storage | Data: 3X 1.92 TB SSD RAID 0 |
| OS Storage | OS: 1X 1.92 TB SSD |
| Network | Dual 10GBASE-T (RJ45) |
RStudio Server and JupyterHub
We have a dedicated server machine for RStudio and JupyterHub. Students taking a CMC course using R or Python, and participating in a QCL workshop may get access to the computational resources. To get access to the RStudio Server and/or JupyterHub, please contact qcl@cmc.edu for your account activated. The RStudio Server and JupyterHub are equipped with 2 CPUs (32 cores) and 512 GB of memory. So, it will perform better than your laptop or personal computer (for sure!) for those who need large memory space.