Case Study: InstaDeep achieves scalable high-performance AI storage with Excelero NVMesh

Instadeep™ Powers Ai As A Service With Shared NVME

InstaDeep, a global AI innovator offering AI as a Service, required a data center infrastructure for its GPU-based systems that could scale modularly, meet flexible performance demands, and maximize GPU utilization for the highest ROI. The challenge was that the local storage on their NVIDIA DGX systems was too limited in capacity and performance, hindering AI workloads that require tens to hundreds of terabytes of data.

Excelero addressed this with its NVMesh software deployed on Boston Flash-IO Talyn servers, creating a scalable, high-performance pool of NVMe storage. This solution provided the GPU servers with remote storage access that performed as if it were local, eliminating data bottlenecks. The results ensured full utilization of the expensive GPU processing power, delivering equal or better performance than local cache and a significantly improved return on investment for InstaDeep's AI service.

InstaDeep