Carina Resources
Overview
The Carina platform offers access to Stanford-managed compute resources for research projects using high-risk data.
The platform is designed for interactive data analytics and offers access to Stanford-managed high-memory servers and GPU-enabled servers.
Disaster recovery and long-term access to data are provided via a combination of on-premise and cloud archive solutions.
A typical suite of analytics tools such as Jupyter, R, Python, and SAS are available.
All available servers can also run traditional jobs through the Slurm scheduler.
Collaboration with other researchers within Stanford is controlled via Stanford workgroups.
The On-Premise vs. Cloud page has some more detailed information about the differences between the Carina On-Prem and Nero GCP environments.
Hardware Technical Specifications
| Type | Quantity | Details |
|---|---|---|
| Login nodes | 2 | |
| Compute nodes | 5 | Dual Core Intel Xeon Gold 6130 (16C 2.1GHz), 370 GB RAM |
| GPU nodes | 2 | Dual Core Intel Xeon Gold 6130 (16C 2.1GHz), 370 GB RAM, Dual P100 GPUs (16GB) |
| GPU nodes | 6 | Dual Core Intel Xeon Silver 4114 (10C 2.2GHz), 250 GB RAM, Quad NVIDIA Tesla V100 |
| GPU nodes | 6 | Dual Core Intel Xeon Gold 6330 (56C 2GHz), 250 GB RAM, Quad NVIDIA A100 |
| Type | Quantity | Details |
|---|---|---|
| Compute nodes | 1 | Dual Core Intel Xeon Gold 6330 (56C 2GHz), 190 GB RAM |
| GPU nodes | 2 | Dual Core Intel Xeon Gold 6330 (28C 2GHz), 450 GB RAM, Quad NVIDIA A100 (40GB) |
| GPU nodes | 1 | Dual Core Intel Xeon Gold 6330 (28C 2GHz), 450 GB RAM, Quad NVIDIA A100 (80GB) |
| GPU nodes | 1 | Dual Core Intel Xeon Gold 6226 (12C 2.7GHz), 1450 GB RAM, 8 NVIDIA Tesla V100 (32GB) |
| GPU nodes | 1 | Dual Core Intel Xeon Gold 6258R (28C 2.7GHz), 700 GB RAM, 8 NVIDIA Tesla V100 (32GB) |
| GPU nodes | 1 | Dual Intel Xeon Platinum 8480+ (56C 2GHz), 940 GB RAM, 8 NVIDIA H100 (80GB) |
| Type | Quantity | Details |
|---|---|---|
| OS | Ubuntu 20.04 | |
| Total Cores | 1,264 | across 6 CPU models |
| Total GPUs | 88 | across 5 GPU models |
| Ethernet Switches | 9 | |
| Storage | 4.5PB | research data storage |
Requesting a Carina On-Premise Project
1. Take a look at the Carina versus Nero GCP cheat sheet to identify whether you need an on-premise solution, a cloud solution or both.
2. Request a Carina On-Premise project using this ServiceNow Carina On-Prem form. Carina On-Prem projects are exclusively designed for projects using high-risk data. Per Stanford University's guidance, a Data Risk Assessment may be required as part of your work with high-risk data. Find out more here.
3. Wait for an email notification confirming your Carina resources are ready.
