Case Study: National Centre for Nuclear Research achieves sub-microsecond, wire-speed supercomputing performance with Arista Networks

National Centre for Nuclear Research - Customer Case Study

The National Centre for Nuclear Research (NCNR) in Poland, home to the Świerk Computing Centre supercomputer, needed a core network refresh to support a major compute upgrade and future expansion. The centre required very high per‑link bandwidth, ultra‑low latency and scalable architecture (MLAG/spine‑and‑leaf) to avoid bottlenecks for HPC workloads and to meet stringent performance and energy targets.

Arista delivered a compact spine‑and‑leaf solution based on Arista 7050 switches and EOS (with Coraid EtherDrive storage), implementing MLAG and in‑switch application scripting. The deployment achieved wire‑speed switching with port‑to‑port latency under 1 µs and 320 Gbps uplinks, doubled effective bandwidth via MLAG, up to three‑times lower energy use versus competitors, and reduced OPEX and management overhead. Combined with a compute upgrade, the centre’s theoretical performance jumped from 17 to 281 teraflops (with 500 TFLOPS projected), enabling full utilization of the HPC cluster.

National Centre for Nuclear Research

Adam Padee

Head of computational infrastructure