NetApp this week introduced Novus, a new storage architecture it says is built for zettascale data. The company claims the architecture can store zettabytes of data and deliver 100 TB/s of aggregate bandwidth over NFS and parallel NFS.
According to NetApp, that bandwidth is enough to keep hundreds of thousands of GPUs productive simultaneously. The architecture is intended to eliminate metadata bottlenecks, which often limit storage performance in large-scale AI workloads. The announcement references "AI Factory" as a target context, pointing to large AI data centers.
This is a single vendor announcement, so the figures and capabilities are NetApp's claims rather than independently verified results. HPCwire reports the specifications as presented, and no third-party benchmarks or comparisons are included in the source.