Persistent Storage

Storage for your ML workloads

Persistent storage that stays with your data. Keep datasets and checkpoints available across pod restarts.

Two storage options

Block storage for fast single-pod access, or NFS for sharing across pods.

Fast I/O

Block Storage

Persistent block storage that attaches directly to your pod. Ideal for checkpoints and workloads requiring fast disk access.

Fast read/write speeds

Attaches to single pod

Persists across restarts

Good for:

Model checkpointsLocal cachingTraining state

Shared Access

NFS Storage

Network file storage mountable to multiple pods simultaneously. Perfect for sharing datasets and models across training jobs.

Mount to multiple pods

Shared file access

Flexible sizing

Good for:

Shared datasetsModel weightsCommon files

What you get

Storage designed for ML workflows with the features you need.

Fast Storage

SSD-backed storage optimized for ML workloads. Fast enough for training and inference.

Persistent

Data persists when pods stop. Resume training without re-downloading datasets.

Flexible Sizes

Choose the storage size you need. Scale up as your datasets grow.

What you can store

From raw datasets to trained models, keep everything organized and accessible.

Datasets

Store training data that persists across sessions.

Checkpoints

Save training progress to resume later.

Model Weights

Keep models ready for fast loading.

Artifacts

Store logs and outputs from experiments.

Simple to use

Create a volume, attach it to your pod, and start using it. Your data stays available even when pods restart.

Mount to pods

Serverless compatible

Data persists

Pay per GB

Volumes

Active

checkpoints

Block • 100 GB

Mounted

45 GB used

45%

datasets

NFS • 500 GB

2 pods

280 GB used

56%

Connected to

my-pod

Ready to get started?

Talk to our team to learn how persistent storage can power your ML workflows.