Case Study: University of Oxford achieves more reliable research storage with WEKA NeuralMesh

University of Oxford modernizes research storage with Weka across 15,000–17,000 compute cores

The University of Oxford's Advanced Research Computing (ARC) service faced challenges with its high-performance scratch storage, including poor serviceability and the risk that a single component failure could compromise its entire shared infrastructure. Their existing system also struggled to provide a single, high-performance namespace across their two separate compute clusters running on InfiniBand and Ethernet fabrics. To build a more reliable foundation for its research community, the university selected the WEKA NeuralMesh platform.

WEKA implemented its NeuralMesh software, which created a single, distributed namespace that natively supports both InfiniBand and Ethernet fabrics. This solution eliminated the blast-radius risk, ensuring a rogue job could no longer take down the entire storage system. The result for the University of Oxford was a significant expansion of storage capacity, a drastic reduction in daily management overhead, and a reliable resource that allows researchers to focus on achieving their goals and publishing significant work.

University of Oxford