NetApp AFX

NetApp AFX is a disaggregated, high-performance storage system built specifically for enterprise AI workloads. It separates the compute layer (storage controllers) from the capacity layer (NVMe enclosures), letting you scale performance and capacity independently. Built on NetApp ONTAP, it delivers extreme throughput, exascale capacity, and enterprise-grade security without the complexity of traditional parallel file systems or proprietary clients.

Traditional all-flash systems couple compute and storage tightly, forcing you to scale both together even when you only need one. AFX's disaggregated architecture breaks that dependency. You can add storage controllers to boost performance or disk enclosures to expand capacity, each with linear gains. This also means nondisruptive upgrades and expansions, no parallel file system outages required.

AFX is purpose-built for AI and ML workloads, including large language model training, deep learning, RAG-based and KV cache inference, and agentic AI. It's also well-suited for high-performance technical computing like seismic analysis and computational fluid dynamics. Because it runs ONTAP, it operates alongside your existing virtualization, database, and enterprise application workloads on a single unified platform.

Yes. AFX is certified for NVIDIA DGX SuperPOD and has been jointly validated by NetApp and NVIDIA. The solution uses pNFS with session trunking, NFS/RDMA for GPU Direct Storage, and FlexGroup volumes with Advanced Capacity Balancing to deliver the throughput required for multimodal training and inference at the largest scales.

AFX supports pNFS, NFS, NFS/RDMA, SMB, and S3. There are no proprietary clients required. Standard protocols mean all your applications can use AFX immediately, without custom software installations that introduce instability or security risk.

AFX uses ONTAP's AI-powered Autonomous Ransomware Protection (ARP) to detect threats in real time. Tamper-proof Snapshot copies prevent attackers from deleting backups, and synchronous zero-data-loss replication keeps a clean copy available for immediate recovery.

Yes. NetApp Keystone delivers AFX as a consumption-based storage-as-a-service subscription. You get AI-optimized performance, independent capacity scaling on demand, predictable billing, and burst capacity at no extra cost. Integrated management through NetApp Console and DII automates operations so your team stays focused on business outcomes, not infrastructure upkeep.

AFX is built on ONTAP, the only enterprise storage OS with first-party, native integrations on AWS, Azure, and Google Cloud. NetApp SnapMirror handles efficient, automated data replication between on-premises and cloud. FlexCache enables ultra-efficient, real-time data distribution across geographies and clouds, so AI workloads access exactly the data they need, instantly and securely, without duplicating full datasets.

An AFX cluster can scale to more than one exabyte of effective capacity with up to 128 storage controllers and 52 NX224 NVMe enclosures. This is managed as a single global namespace, which simplifies operations significantly compared to managing multiple storage silos.

AIDE is an ONTAP-integrated data service that accelerates AI data pipelines end to end. It provides a global, real-time metadata catalog for fast data discovery and curation, automates change detection and data synchronization, enables in-place vectorization for GenAI workloads powered by NVIDIA NIM, and enforces built-in governance guardrails throughout the AI lifecycle. Together, AFX and AIDE form a complete enterprise-grade solution for operationalizing AI at scale.