Data Analysis Cluster "BladeRunner"

The main computing facility of the Open Physics Hub - the HPC cluster "Matrix" - is directly connected with a smaller cluster primarily devoted to post-processing and data analysis activities, the BladeRunner cluster, equipped with a heterogeneous architecture of computing nodes with different numbers of cores, memory, and network connections, and with a dedicated disk storage space of 43 TB backed-up with RAID-5 protocols.

The different computing nodes of the BladeRunner cluster have been tailored for different types of post-processing and data analysis jobs that are handled and prioritized by a dedicated job scheduler (SLURM). All the nodes have x86_64 instruction set and their main specifications are as follows:

Infiniband-100 connected nodes:

• 2 nodes with 24 virtual cores and 64 GB RAM
• 2 nodes with 32 virtual cores and 64 GB RAM
• 1 node with 32 virtual cores and 128 GB RAM
• 2 nodes with 16 virtual cores and 24 GB RAM

Ethernet connected nodes

• 2 nodes with 12 virtual cores and 16 GB RAM
• 2 nodes with 32 virtual cores and 64 GB RAM

The BladeRunner storage space also features a web-accessible public repository for the publication and dissemination of data products.