MayaNAS · on Google Cloud

MayaNAS on Google Cloud.

One engine on Cloud Storage — a parallel storage engine with Lustre to feed your GPUs, and enterprise NAS for everything else. Deploy from Google Cloud Marketplace into your own project, billed through your Google Cloud account.

Choose your deployment

One engine, two halves — each on Google Cloud Marketplace.

Both run in your project with your Cloud Storage buckets as the capacity tier. Pick by what your clients need.

Parallel at scale · 5-day trial

Lustre on Cloud Storage

Choose this when GPUs need a shared parallel filesystem — training data, checkpoints, inference caches.

  • 1 MGS/MDT + 2 OSTs with active-active failover
  • Stock in-kernel Lustre client; on GKE, Google's own Lustre CSI driver mounts it
  • MLPerf® Storage v3.0 results served from Cloud Storage
Enterprise NAS

Enterprise NAS on Cloud Storage

Choose this for shared file and block storage — home directories, tools and notebooks, media projects, Windows shares, application volumes.

  • NFS and SMB shares; iSCSI and NVMe/TCP block volumes
  • Snapshots, replication, compression and checksums
  • Active-active HA pair; capacity grows at Cloud Storage prices

Need Lustre and NAS on the same data? It is one engine — talk to us.

Our MLCommons-verified MLPerf® Storage v3.0 results on Google Cloud
14.43 GiB/s
checkpoint write, Llama 3 8B, one client
32.42 GiB/s
checkpoint write, Llama 3 70B, two clients
GCS
every result served from Google Cloud Storage
Built for Google Cloud

Runs the way your project already does.

Billed through Google Cloud

Subscribe on Google Cloud Marketplace and pay on the Google Cloud account you already have — no separate procurement.

Your buckets are the capacity

Data lives in Cloud Storage buckets in your own project and region. Capacity grows at bucket prices, with nothing to pre-provision.

Ready for GKE

Google's bundled Lustre CSI driver mounts MayaNAS Lustre with a static volume, and NFS works with standard Kubernetes volumes — nothing of ours on the cluster.

Inside your VPC

Storage nodes deploy into your project and VPC network on private addresses, next to the Compute Engine and GPU instances they serve.

Feed GPUs on Google Cloud, at object-storage cost.

Start with the Lustre trial, or talk to us about sizing, multi-protocol deployments and running across clouds.

MLPerf® Storage v3.0, Closed division, submitted by ZettaLane Systems on Google Cloud. Entries 3.0-0137 (Llama 3 8B checkpointing) and 3.0-0139 (Llama 3 70B checkpointing). Result verified by MLCommons Association. Retrieved from mlcommons.org/benchmarks/storage/ on 7 September 2026. The MLPerf name and logo are registered and unregistered trademarks of MLCommons Association in the United States and other countries. All rights reserved. Unauthorized use strictly prohibited. See www.mlcommons.org for more information.