MayaNAS runs enterprise NAS and the Lustre parallel file system from one engine on standard Google Cloud infrastructure, with bulk data in Google Cloud Storage.
ZettaLane Systems, LLC, a Select Google Cloud Technology Partner, today announced that MayaNAS, the only MLPerf® Storage v3.0 submission verified running the Lustre parallel file system directly on cloud object storage, is available on Google Cloud Marketplace. MayaNAS presents standard enterprise NAS (NFS, and SMB with Active Directory) alongside the parallel file systems Lustre and pNFS, with bulk data stored in Google Cloud Storage.
ZettaLane can now securely scale and support enterprises using MayaNAS to deliver high-performance, shared storage for demanding workloads, including AI training.”
— Dai Vu, Managing Director, Marketplace & ISV GTM Programs, Google Cloud
In MLPerf® Storage v3.0, run entirely on standard Compute Engine instances with Google Cloud Storage, MayaNAS sustained 32.42 GiB/s checkpoint write and 21.20 GiB/s read on the Llama 3 70B workload with two client nodes and four storage nodes. On the 3D U-Net training workload, a single client node kept three simulated NVIDIA B200 accelerators about 92% busy. The result shows that one file system on standard Google Cloud infrastructure can serve both enterprise file workloads and the most demanding parallel ones, with Google Cloud Storage as the storage layer.
Also Read: CIO Influence Interview with Srinivasan Panchapakesan, Chief Platform Officer at Hexaware
The approach has drawn independent attention in the storage press. Blocks & Files wrote: “Here’s an interesting idea; ZettaLane has built MayaNAS, a Lustre parallel filesystem, with NFS, pNFS and SMB support, based on cloud object storage.” Storage Newsletter called it a “pretty innovative approach.”
Most cloud NAS stops where HPC and AI begin. On Google Cloud, MayaNAS lets teams scale from everyday file shares to GPU-class parallel throughput without deploying a second storage system. Bulk data lives permanently in Google Cloud Storage, so there are no tiering or staging jobs between a fast tier and the bucket. Capacity grows with the bucket, and teams pay for performance only while they need it: serving nodes scale up for a run and back down afterward while the data stays in place. The file system, buckets and encryption keys remain inside the customer’s own project, supporting data-residency and sovereignty requirements. MayaNAS runs active-active for high availability and uses the standard open-source clients already in Linux, with no proprietary software to install.
Beyond AI, the same parallel throughput serves genomics and life-sciences pipelines, electronic design automation, media rendering, financial risk analysis and seismic processing. MayaNAS is built for enterprise IT, research institutions, regulated industries and sovereign AI initiatives that need parallel file performance at object-storage scale while keeping full control of their data.
“Cloud object storage scales without limits and handles durability on its own. We built MayaNAS to use it directly as the storage layer beneath a parallel file system, so teams get the enterprise file services they already run and the parallel Lustre and pNFS throughput their GPUs need, with no disks, RAID or separate fast tier to manage,” said Supramani Sammandam, Founder of ZettaLane Systems. “MLPerf® Storage v3.0 showed that standard Compute Engine instances and Google Cloud Storage make a powerful combination for serving the AI data pipeline, from training and checkpointing to inference cache, with the data staying in the customer’s own project.”
“Bringing ZettaLane Systems to Google Cloud Marketplace will help customers quickly deploy, manage, and grow the company’s MayaNAS storage platform on Google Cloud’s trusted, global infrastructure,” said Dai Vu, Managing Director, Marketplace & ISV GTM Programs at Google Cloud. “ZettaLane can now securely scale and support enterprises using MayaNAS to deliver high-performance, shared storage for demanding workloads, including AI training.”
Customers can adopt MayaNAS the way they already buy cloud: pay-as-you-go on Google Cloud Marketplace with a f*********, deployed in minutes into their own VPC and billed through their Google Cloud account, with eligible purchases counting toward existing Google Cloud commitments; through private offers for negotiated terms, directly or via Google Cloud reseller partners; or as code, using ZettaLane’s open-source Terraform modules, which deploy the same Marketplace image and also cover pNFS, for teams that run their infrastructure as code. To start an evaluation or plan a deployment, contact sales@zettalane.com. MayaNAS is part of ZettaLane’s storage family for AI data and agentic workloads. Learn more at zettalane.com.
Catch more CIO Insights: The Overlooked Security Risk Behind Enterprise AI Rollouts
[To share your insights with us, please write to psen@itechseries.com ]
