Unlike AFF and FAS storage systems, the AFX storage and compute elements of the cluster are decoupled. AFX controllers are not restricted to accessing a limited set of local disks.

The storage administrator no longer needs to manage the aggregates and RAID groups. AFX handles these tasks which simplifies administration and provides an opportunity for nonspecialists to manage their data.

Decoupling the AFX compute and storage, along with simplified storage management, results in a single pool of storage. This Storage Availability Zone (SAZ) is available to all the controller nodes in an AFX cluster. See FAQ for AFX storage systems for more details.

AFX is built for the high performance needed by specialized AI/ML applications. It can make more efficient use of modern hardware to provide high and sustained bandwidth with ultra‑low latency.

The single storage pool, along with other features including automatic storage management, provides a much simpler administrative and operational model.

The set of controller nodes and storage shelves in a cluster can be independently expanded based on the application needs. There are limited administration requirements and disruptions when adding controllers or shelves.

Volumes can be moved among the cluster controller nodes non-disruptively without actually copying data. Metadata directory and index pointers are updated instead. This dramatically improves performance during typical administrative procedures as well as recovery in failure scenarios.