MayaNAS serves Lustre parallel file and familiar NAS — NFS, pNFS, SMB — directly out of your cloud's object storage. Training data streams at GPU speed, checkpoints land at tens of gigabytes per second, and capacity costs what your buckets cost. Validated by MLPerf Storage benchmarks.
One namespace, hundreds of compute clients, striped throughput that scales with the cluster — the same parallel filesystem the world's largest GPU clusters train on, with standard clients.
NFS, pNFS, and SMB from the same system for everything around training — notebooks, tools, home directories, render farms. Full POSIX, nothing to install.
Object storage is the substrate, not a backup target — data is read and written in place. No staging pipeline, no cache tier to size and warm, and capacity grows at bucket prices.
Absorbs the burst writes of LLM checkpointing at tens of gigabytes per second, then drains to object storage — training resumes fast, and every checkpoint lands durably.
Active-active server pairs with automatic failover. Clients reconnect and carry on — no rerun of the training job, no manual recovery steps.
Deploy from your cloud marketplace into your own account and VPC. Sized configurations from entry to ultra, ready in minutes, billed through your cloud agreement.
Deploy with terraform into your own account and VPC — sized configurations, ready in minutes, billed through the cloud agreement you already have. Prefer a guided first run? Talk to us.
Terraform deployment into your AWS account — S3 as the object tier.
Terraform deployment into your GCP project — GCS as the object tier.
Terraform deployment into your Azure subscription — Blob as the object tier.
Self-serve with terraform from your cloud marketplace — ready for your most demanding training runs.