NetApp’s Novus tackles metadata bottlenecks in AI data infrastructure
SiliconANGLE Sloane Kali Faye ● Covered by 2 sources
NetApp says its Novus system is built to fix AI data bottlenecks, not just speed storage. The big trick is separating metadata from data so enterprise AI can scale without ripping out old systems.
Based on reporting by SiliconANGLE, Sloane Kali Faye — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
NetApp is pitching Novus as more than another faster box for AI. The company wants to solve a mess that shows up before the model even gets to work: data spread across systems, rules applied by hand, and metadata that becomes the choke point when AI traffic starts piling up.
At NetApp INSIGHT, chief product officer Syam Nair argued that plenty of companies already have the parts they need — data applications, models, investment — but still fail to see returns. His answer is a unified storage layer that spans the infrastructure instead of forcing customers to stitch everything together by hand. That is the pitch, at least: less fragmentation, less waste, more usable data.
The Novus storage architecture is built around scale. NetApp says it exceeds 100TB/s of aggregate throughput, while its Data Director keeps metadata separate from the stored data so both can grow on their own. Gunna Marripudi, vice president of product management for Novus, said the point of that split is concurrency: metadata access has to keep up if the architecture is going to hold together. Novus also gives applications a single view of files across multiple ONTAP storage clusters, rather than making them hit each one separately.
But raw speed is only part of the story. Nair said NetApp’s AI Data Engine is meant to discover, classify and vectorize enterprise data, and reduce a preparation job that often takes six to nine months of engineering work. His warning was blunt: fast access is necessary, but not enough if the data is still messy or unprotected. Otherwise, AI can produce quick answers that are still wrong, or worse, compromised.
NetApp is also leaning on open standards to make this less disruptive. The planned acquisition of PEAK:AIO Ltd. is expected to add parallel file system technology based on pNFS, which Marripudi said already has a client in Linux kernels released after 2018. The message underneath all of it is simple: NetApp wants to sell AI infrastructure without asking customers to throw out what they already run.
My take — AI-written commentary, not fact-checked reporting
NetApp is smart to talk about metadata instead of waving at GPUs like that solves everything. AI shops keep buying speed and then act shocked when the real bottleneck is the swamp underneath. Open standards help here because nobody needs another heroic rip-and-replace project dressed up as innovation.
Read more about this at: SiliconANGLE