Google Cloud Managed Lustre delivers a high-performance, fully managed parallel file system optimized for AI and HPC applications. With multi-petabyte-scale capacity and up to 10 TBps throughput, Managed Lustre facilitates the migration of demanding workloads to the cloud.
Highlights
- Developed in collaboration with DDN, the leading provider and primary maintainer of the Lustre file system.
- Supports extreme IOPs and low latency.
- POSIX compliance ensures compatibility with existing applications, tools, and workflows.
- Integrated with Compute Engine and Google Kubernetes Engine, allowing you to mount high-performance storage to your virtual machines and containerized applications.
- Provides high-speed and fully managed transfers into and out of Cloud Storage.
Specifications
Up to 5,600 write IOPS per TiB
Per 72 GBps of throughput, up to:
- 410,000 stats per second
- 115,000 creates per second
- 95,000 deletes per second
Up to a maximum of 22 times the base value.
For more information about performance, see Performance considerations . For more information about quotas and limits, see Quotas and limits .
Use Cases
-
High-performance computing: Managed Lustre excels in HPC environments where multiple compute nodes need fast and consistent access to shared data for simulations, modeling, and analysis.
-
Machine learning: Managed Lustre can handle the large datasets and high throughput requirements of machine learning workloads, enabling efficient training and inference.
Data protection
Managed Lustre encrypts data at rest and data in transit.
Encryption at rest
All data stored in a Managed Lustre file system is encrypted at rest.
For more information, see Encryption at rest .
Encryption in transit
While the Lustre protocol itself doesn't encrypt traffic between clients and servers, all in-transit data to and within Google Cloud is encrypted. This includes data transfers between Managed Lustre and Cloud Storage.
For more information, see Encryption in transit .
Maintenance and availability
Managed Lustre is designed for high availability. However, certain configurations that provide extreme performance and scale may require planned downtime of up to 4 hours for periodic host maintenance. Google coordinates these maintenance events with you in advance.
For more information, see Troubleshooting and known issues .
Pricing
See the Pricing page for details.
