High-performance computing service

The ModES team contributed to the funding of a local computing cluster, along with storage services and pre-/post-processing nodes.
With the support of:

Responsible : Aymeric BLONDEL
The Pekka local computing resources complement the regional and national offerings. They are dedicated to specific use cases and are managed by the laboratory’s high-performance computing service, which provides researchers with the most suitable solution for their needs.
At the local level, these resources are primarily used for:
- Prototyping,
- Pre- and post-processing of calculations,
- High-memory requirements,
- Calculations requiring fast local storage,
- Non-restartable calculations with non-standard runtime needs.
For example, the Pekka cluster has enabled highly resource-intensive calculations, such as the series of publications contributing to Quest, whose goal is to provide reference vertical excitation energies for small-and medium-sized molecules. This data allows for the evaluation and comparison of the accuracy of computational quantum chemistry methods.
Pekka cluster:
Last updated: June 15, 2026
Production deployment date: 2014 (ClusterVision solution)
Full reinstallation: 2022 (with open-source solutions)
Cooling: Modular Liquid Cooling Unit (LCP)
Operating System: AlmaLinux
Job Manager: Slurm
| CPU series | Nb Nodes | Cores/node | Ram/node | Local disk /work | Walltime max |
| Intel E5-2670 | 4 | 16 | 128 Go | 500 Go | 240 H |
| Intel Xeon 4114 | 3 | 20 | 96 Go | 900 Go | 48 H |
| Intel E5-2697 v3 | 3 | 28 | 192 and one at 256 Go | 4 To | 2160 H |
| AMD EPYC 7443 | 2 | 48 | 2 To | 6.5 To | 2160 H |
| AMD EPYC 9354 | 2 | 64 | 1..5 To | 6.5 To | 2160 H |
| Intel Xeon(R) Gold 5320 | 1 | 52 | 4 To | 12 To | 2160 H |
Visualization and Pre-/Post-Processing: Features
- Stations 1 & 3:
Prototyping of applications, containerization, creation of computational workflows, inference, and model training—prior to potential transfer to the regional mesocenter.
These two stations can also launch calculations on the Pekka cluster nodes. - Station 2:
Hosting of analysis or computational data processing applications.
Example: Hosting of the Cambridge Structural Database (CSD) and its associated software.
Logiciels de chimie quantique:
Last updated: 15/06/2026
- CFOUR
- Dalton
- MRCC
- OpenMolcas
- Psi4
- DFTB+
- ORCA
- Q-Chem
- Turbomole
- Gaussian
- fcClasses
- GROMACS
- SHARC
- OpenMolcas
* Some software may not be installed simultaneously on both platforms, GLiCID and Pekka. Availability should therefore be checked on a case-by-case basis.
Storage
The MODES team has access to 300 TB of warm storage provided by GLiCID. A backup solution for a portion of the warm and cold data generated from calculations is managed by the CEISAM High-Performance Computing service.
Documentation is available on the intranet.
