Lustre File System
Overview
Lustre is an open-source, parallel distributed file system designed for high-performance computing (HPC) environments. It provides scalable storage for large clusters, supporting POSIX compliance, high I/O throughput, and management of massive datasets. Lustre separates metadata and data handling across dedicated servers, enabling it to scale to thousands of clients and petabytes of storage.
History
The Lustre file system originated as a research project in 1999 at Carnegie Mellon University, led by Peter J. Braam. It evolved from earlier work on the InterMezzo and Coda file systems, focusing on object-based storage for Linux clusters. In 2001, Braam founded Cluster File Systems (CFS) to commercialize the technology, with funding from the U.S. Department of Energy's Accelerated Strategic Computing Initiative (ASCI) Path Forward project, in collaboration with Hewlett-Packard and Intel.
The first production deployment occurred in March 2003 on the MCR Linux Cluster at Lawrence Livermore National Laboratory, then ranked third on the TOP500 supercomputer list. Lustre 1.0.0 was released in December 2003, introducing basic failover and recovery features. By 2006, it powered 10 of the top 30 supercomputers worldwide.
Sun Microsystems acquired CFS in 2007, integrating Lustre into its HPC portfolio and exploring merges with ZFS. After Oracle acquired Sun in 2010 and halted Lustre 2.x development, the community formed the Open Scalable File Systems (OpenSFS) organization to sustain open-source efforts. Intel acquired Whamcloud (a key contributor) in 2012 to advance exascale computing. In 2018, DataDirect Networks (DDN) acquired Lustre assets from Intel, boosting development. The Lustre trademark transferred to OpenSFS in 2019.
Key releases include:
- Lustre 1.2.0 (2004): Linux kernel 2.6 support and client-side caching.
- Lustre 1.4.0 (2004): InfiniBand integration.
- Lustre 1.6.0 (2007): Dynamic OST addition.
- Lustre 2.0 (2010): Major architecture overhaul.
- Lustre 2.1 (2011): Community-led, Red Hat 6 support.
- Lustre 2.4 (2013): Distributed Namespace Environment (DNE) for metadata scaling and ZFS backend.
- Lustre 2.5 (2013): Hierarchical Storage Management (HSM) for tiered storage.
By 2012, Lustre ran on 59 of the top 100 supercomputers.
Major Accomplishments
Lustre has become a cornerstone of HPC, powering over 60% of the TOP500 supercomputers, including more than half of the top 10. Notable deployments include:
- Frontier (2022, world's No. 1 TOP500, 700 PB capacity, exascale).
- Fugaku, Titan, and Sequoia.
It supports clusters with tens of thousands of clients, hundreds of petabytes of storage, and aggregate I/O up to tens of terabytes per second. Accomplishments span industries like scientific simulation, AI/machine learning, oil and gas, life sciences, and finance. Innovations such as DNE (phased rollout 2013-2016) enable metadata scaling across multiple servers, while HSM allows seamless data tiering to archival storage. Lustre's object-based design and RDMA support (via LNet for InfiniBand/Ethernet) deliver superior performance over traditional file systems like NFS or GPFS in parallel environments.
Current State
As of September 2025, Lustre's latest general availability (GA) release is 2.15.7, declared in June 2025. This long-term support (LTS) version builds on 2.15.0 (2023), emphasizing stability, interoperability, and modern infrastructure support. Key features in the 2.15 series include:
- Client Logical Metadata Volume (LMV) with round-robin and space-balanced directory layouts for efficient use of multiple Metadata Targets (MDTs).
- IPv6 support (fully merged in 2.15.x).
- RHEL 9.x server and client support (up to 9.6), plus SLES 15 SP6 client compatibility.
- Enhanced security with fscrypt integration (2024).
- Ongoing metadata optimizations, such as Lazy Size on MDS.
Lustre remains dominant in HPC and is expanding into AI/ML workloads, with managed services like AWS FSx for Lustre defaulting to 2.15 for new file systems. Developments include the Integrated Manager for Lustre (2021) for easier deployment and efforts to upstream components into the Linux kernel (discussed at 2025 Linux Storage Summit). It supports backends like ldiskfs (ext4) and ZFS, with elasticity for dynamic resource addition. Future LTS is 2.15.8, focusing on upgrades from prior 2.15.x releases.