Vega cluster

Vega is the CECI cluster installed at ULB. It features 42 compute nodes and 2 GPU nodes (NVIDIA Tesla M2090), interconnected with an Infiniband QDR network, and 70 TB of high performance storage under GPFS.
Each compute node has 4 AMD Opteron 6272 2.1GHz (16 cores), 256 GB of RAM and 250 GB of local disk space. In total, Vega has 2752 cores and 11TB of RAM. The Linpack benchmark gives 16.2 TFLOPS of aggregated peak computing power performance. The nodes density makes Vega suitable for many-cores (SMP and MPI) and many single core jobs. With 256 GB of RAM per node, memory intensive jobs should easily scale on Vega. Given the CPU type, multi-core integer-based calculations will be highly efficient while jobs requiring top performance on floating point calculations should go to more specialised clusters. The storage solution allows to run heavy disk IO jobs.
See the specifications page for more details.


There are two sources of documentation for Vega:
HowTos: a set of pages describing how to perform common operations on the Vega cluster.
    Topics are: Q&A: questions and answers.