Google Cloud Managed Lustre
cloud.google.com
- Linux
- Web
- API
- Self-hosted
What Google Cloud Managed Lustre is
Google Cloud Managed Lustre is a managed parallel file system for AI and high-performance computing workloads. It supports throughput up to 10 TBps and storage capacity up to 84,016,000 GiB (80.1 PiB), with POSIX compliance for compatibility with existing applications, tools, and workflows. Instances can be mounted to Compute Engine VMs and Google Kubernetes Engine clusters. The service imports data from and exports data to Cloud Storage, including incremental transfers of new or changed files. Data is encrypted at rest and in transit to and within Google Cloud, and per-instance policies allow customers to set maintenance windows and exclusion periods. Documented workloads include HPC simulations, modeling and analysis, and machine-learning training and inference. Capacity is provisioned, and unused provisioned capacity is billed. Pricing is listed from $0.06/mo. Limits include a maximum file size of 0.5 PiB, 20 instances per project, and one active transfer operation per instance.
Who it suits
It suits teams running HPC or machine-learning workloads that need shared file-system access from Google Cloud compute. It is also relevant to administrators whose workflows depend on POSIX compatibility and Cloud Storage transfers.
Good to know, and what to watch
- Throughput up to 10 TBps
- Supports up to 84,016,000 GiB of storage
- Mounts to Compute Engine VMs and GKE clusters
- Encrypts data at rest and in transit
- Incremental Cloud Storage transfers are supported
- No free plan or free trial
- Unused provisioned capacity is billed
- One active transfer per instance
- Maximum file size is 0.5 PiB
Plans
Every plan and priceDynamic (25 MB/s/TB)
Dynamic (25 MB/s/TiB)
125 MB/s/TiB
250 MB/s/TiB
The verdict
Managed Lustre combines high throughput, POSIX compatibility, and integrations with Google Cloud compute and storage. Account for provisioned-capacity billing and the file-size and transfer-operation limits when planning workloads.
What the maker publishes
- API surface
- Google publishes a v1 REST API with location, instance and operation resources and a lustre.googleapis.com endpoint. source
- Authentication
- Programmatic access uses Google Cloud authentication and IAM authorization, with workload identity federation recommended outside Google Cloud. source
- Availability target
- The Managed Lustre SLA specifies at least 99.9% monthly uptime for an instance. source
- Capacity
- The documentation lists capacity from 9,000 GiB to 12,240,000 GiB and maximum file size of 0.5 PiB. source
- Capacity limits
- The maximum file size is 0.5 PiB, and the maximum number of instances per project is 20. source
- Client operating systems
- The Compute Engine guide supports HPC Rocky Linux 8, Rocky Linux 9, RHEL 9, Ubuntu 22.04 and Ubuntu 24.04, with Ubuntu 24.04 limited to Arm VMs. source
- Client requirement
- Lustre client packages are kernel-version specific on Ubuntu, so kernel upgrades can require reinstalling matching packages. source
- Cloud Storage integration
- Managed Lustre can import data from and export data to Cloud Storage, with incremental transfers that copy new or changed files. source
- Cloud Storage integration
- Managed Lustre provides managed high-speed transfers into and out of Cloud Storage. source
- Company founded
- Gartner’s company details list Google as founded in 1998 and headquartered in Mountain View, California. source
- Company headquarters and size
- Google’s LinkedIn profile lists Mountain View, California headquarters and 10,001-plus employees. source
- Company legal name
- Crunchbase identifies the maker as Google LLC and lists it as a subsidiary of Alphabet. source
- Compatibility
- POSIX compliance supports compatibility with existing applications, tools, and workflows. source
- Compute integrations
- Instances can be mounted to Compute Engine VMs and Google Kubernetes Engine clusters. source
- Compute integrations
- Instances can be mounted by Compute Engine virtual machines and Google Kubernetes Engine applications. source
- Developer tooling
- Documentation links REST, gcloud lustre commands, Terraform resources and a Cluster Toolkit module. source
- Dynamic tier
- The Dynamic tier adds scalable capacity, a caching layer and cache monitoring for cacheable workloads. source
- Independent technology announcement
- DDN describes Managed Lustre as a zonal service with petabyte-scale capacity, sub-millisecond latency, millions of IOPS and up to 1 TB/s throughput in its 2025 announcement. source
- Intended workloads
- The documentation identifies HPC simulations, modeling and analysis, and machine-learning training and inference as use cases. source
- IOPS
- Published limits reach 5,800 read IOPS and 5,600 write IOPS per TiB. source
- Licence: Lustre
- The community Lustre filesystem is open source; its kernel modules use GPL-2.0 with the Linux syscall exception and other components use compatible per-file licences. source
- Limit: Google Cloud billing complexity
- The same Gartner review said billing complexity causes headaches despite positive overall experience. source
- Limit: project quota
- The quotas page documents a maximum of 20 Managed Lustre instances per project and 20,000 client connections per instance. source
- Limit: zonal placement
- The product is provisioned in an individual zone and the documentation recommends locating it in the same zone as client VMs or GKE clusters. source
Checked against the maker's pages on 15 Oct 2026.