CEDIA's HPC cluster¶
High performance computing infrastructure operated by CEDIA from its Data Center in Cuenca (Ecuador), with ISO/IEC 27001:2022 certification for its information security management system.
Computing capacity¶
| Resource | Detail |
|---|---|
| GPU nodes | 3 NVIDIA DGX A100 nodes, 8 A100 40 GB accelerators each (24 total) |
| CPU | 768 cores per node |
| Memory | ~3 TB RAM per node |
| Network | 40/100 GbE high-performance interconnect |
Storage¶
| Space | Purpose | Persistence |
|---|---|---|
$HOME (~/) |
Personal files, configurations and final results | Permanent; biweekly backups |
/work/$USER (1 TB per user) |
Datasets, active projects and results: ~110 TB parallel high-performance file system. Available on the nodes where your jobs run | Persistent, but not a backup: when your research ends, move what matters to $HOME and clean it up |
$SCRATCH |
Ultra-fast local disks of the node running your job | Temporary: deleted when the job ends |
The recommended workflow is /work → $SCRATCH → processing → /work.
Software¶
Scientific software is delivered compiled and optimized for this hardware and managed with modules: a single command loads the application with all its dependencies. The catalog exceeds 45 applications and includes compilers and libraries (GCC up to 12, CUDA up to 12.8, OpenMPI, OpenBLAS, HDF5, NetCDF), machine learning frameworks (PyTorch, JupyterLab), simulation codes (GROMACS, LAMMPS, Quantum ESPRESSO, SIESTA, ORCA, WRF), geospatial tools (QGIS, GDAL) and bioinformatics tools (BLAST, MMseqs2). See the full application catalog.
Job management¶
Jobs are scheduled with SLURM: you request the partition, cores, memory and GPUs you need, and the scheduler assigns them when available. Queues are organized by usage profile (development, production and large jobs, on CPU and GPU), and visual portal sessions automatically use dedicated resources.
Platform services¶
- Visual access portal (Open OnDemand): Jupyter, RStudio, QGIS and remote desktop with GPU from your browser.
- Account requests: requirements, process and approval of CEDIA user accounts.
- Generative AI pilot: LLMs with an OpenAI-compatible API, RAG and embeddings.
- NOC support Monday through Sunday, 07:00 to 23:00 (contact channels).
Using this cluster in a publication? We appreciate citing the service in the acknowledgments section. See the acceptable use policy.