Overview
Network Volume is a persistent, network-attached storage service for GPU Virtual Machines on FPT AI Factory. Unlike the temporary local disk bundled with a GPU VM, a Network Volume survives VM stop/start cycles and can be mounted by multiple VMs in the same workspace simultaneously.
Supported configurations
| Details | |
|---|---|
| Compatible resources | GPU VMs only |
| Supported GPU types | H100, RTX PRO 6000 (Ha Noi region) |
| Capacity range | 50 GB – 3,000 GB |
| Sharing | Multiple VMs in the same workspace can mount the same volume |
Key concepts
Persistent storage — Data on a Network Volume is preserved when the GPU VM is stopped, restarted, or replaced. This makes it suitable for storing datasets, model checkpoints, and experiment outputs.
Shared access — VMs in the same workspace can mount the same Network Volume at the same time. This allows multiple nodes to read from or write to shared data without copying files between machines.
Region — A Network Volume is created in a specific region and can only be attached to GPU VMs in that same region.
Volume lifecycle — A volume exists independently of any VM. You can detach it from one VM, attach it to another, or leave it unattached. You are billed for the provisioned capacity for as long as the volume exists.