Bacula Systems has expanded its high-performance computing data protection portfolio with new BsnapDiff technologies for ZFS and GPFS, alongside Lustre changelog integration, aiming to dramatically accelerate incremental backups and reduce the strain of protecting massive, file-heavy environments
As data volumes explode across research labs, AI clusters, supercomputing facilities and enterprise high-performance computing environments, backup is becoming one of the most difficult infrastructure challenges to manage at scale. Bacula Systems is betting that the answer lies not in brute-force scanning, but in making backup software smarter about the file systems it protects. The company has now expanded its HPC data protection portfolio with the launch of BsnapDiff technology for ZFS and GPFS environments, alongside a Lustre Changelog integration designed for extreme-scale deployments.
The new capabilities are aimed squarely at one of the most persistent problems in high-performance computing and large archival environments: how to perform incremental backups efficiently when storage systems contain hundreds of millions, or even billions, of files. In such settings, traditional backup approaches can become painfully inefficient, often forcing systems to scan entire directory trees just to identify which files have changed since the previous backup.
That process is not just slow. It also places significant pressure on production storage, drives up metadata activity, extends backup windows and can affect the performance of applications and users sharing the same infrastructure. Bacula’s latest release is intended to address precisely that pain point by shifting away from full file system scans and towards native, file system-aware change detection.
A different approach to incremental backup at scale
Rather than crawling every directory to locate changed files, Bacula’s new BsnapDiff technologies use the underlying capabilities of each file system to identify what has changed more directly. Depending on the environment, that may mean drawing on metadata, snapshots, APIs or changelog events to generate a much more efficient picture of file-level changes.
In practice, the benefit is straightforward: less scanning, less overhead and far faster incremental backups.
“Large-scale infrastructures require data protection technologies that work with the file system rather than against it,” said Jorge Gea, CTO of Bacula Systems. He added that the new BsnapDiff capabilities represent another step in Bacula’s strategy of deep integration with leading high-performance file systems, helping organisations improve backup performance, reduce infrastructure load and align data protection more effectively with modern HPC architectures.
For environments operating at serious scale, that kind of optimisation can have an outsized impact. In many HPC and research systems, only a small percentage of files change between backup cycles, yet conventional backup tools still perform exhaustive scans of the entire file structure. Bacula’s argument is that this is no longer sustainable in an era of rapidly expanding datasets, AI training environments and exascale computing ambitions.
Built for ZFS, GPFS and Lustre-heavy architectures
The new technologies are tailored to the realities of some of the most widely used storage architectures in HPC and research computing. In ZFS environments, Bacula’s BsnapDiff uses zfs diff to identify changes between snapshots. The company says that in infrastructures containing hundreds of millions of files, generating a list of modified files can take less than a minute—compared with the many hours that a conventional full-tree scan may require.
In GPFS environments, Bacula’s approach leverages IBM Storage Scale snapshot and backup APIs to identify changed inodes efficiently, without relying on directory traversal. That reduces unnecessary reads and helps minimise the kind of metadata churn that can slow down large systems.
For Lustre deployments, where snapshots are often less practical at extreme scale, Bacula has introduced a Lustre Changelog Plugin. This consumes and tracks Lustre changelog events to maintain an optimised record of changes, enabling more efficient incremental backups without the heavy operational cost of scanning the entire namespace.
The practical benefits of this file system-aware model are significant. Bacula says the technology can deliver dramatically faster incremental backups, reduced impact on primary storage, improved metadata efficiency, shorter backup windows and better user and application performance. In environments where backup jobs often compete with production workloads, that reduction in overhead can be especially valuable.
Part of a wider HPC strategy
The launch is not an isolated product enhancement but part of Bacula’s broader strategy in the HPC and large-scale storage market. The company has been steadily building out technologies designed for demanding research, AI and enterprise environments, including its HPCAccelerator feature and a growing set of native integrations with high-performance file systems.
Taken together, these investments suggest a clear direction: Bacula wants to position itself not simply as a backup vendor, but as a specialist in protecting some of the most data-intensive infrastructures in the world. That includes supercomputing facilities, scientific research platforms, AI environments and enterprises operating large archival systems where scalability, resilience and operational control are paramount.
As organisations continue to build ever larger data estates, the backup problem becomes less about capacity alone and more about architecture. The challenge is no longer just where to store copies of data, but how to move, index and recover that data without overwhelming the very systems it is meant to protect. Bacula’s latest enhancements are aimed squarely at that challenge.
In a market where AI workloads, scientific datasets and high-performance storage are all growing at breakneck speed, Bacula is making a clear case for backup software that understands the structure of the environment beneath it. If conventional file-by-file scanning increasingly looks like a relic of a smaller-data era, BsnapDiff may be Bacula’s answer to what next-generation data protection needs to look like.
