Introducing Devana¶
Devana is the HPC cluster procured by the Computing Centre SAS and put into operation in Q2 of 2023. It consists of 148 compute nodes, totaling 9,472 compute cores, 38 TB RAM and 32 GPU accelerators. The system's theoretical peak performance is about 800 TFLOP/s.
All nodes are interconnected through a fully non-blocking fat-tree InfiniBand network (100 Gb/s on the universal nodes, 200 Gb/s HDR on the accelerated nodes), and are equipped with Intel® Xeon® Ice Lake architecture processors. There is an accelerated module composed of 8 nodes, each augmented with four NVIDIA® A100 accelerators. Read more in Compute Nodes section.
The cluster runs the Rocky Linux 8.10 operating system (Red Hat Linux family). A large variety of utilities, scientific applications and programming libraries (see sections Software, Compilation, and Environment for more details) maintained by the HPC team is available to the user community. The user software stack is generated using Easybuild and is accessible via environment modules.
There are three shared filesystems available to users on the cluster
nodes. Persistent user and project data are mounted on the /home and
/project mount points, respectively, and temporary calculation
data reside on /scratch mount point. Data stored on
/project mountpoint are shared across the Perun and Devana
clusters. Read the Mountpoints section
for details.
The open-source Slurm workload manager is used for resource and job management. It provides detailed computing-resource allocation and job execution. For more information, see the Job Submission and Devana-specific Partitions section.
