Since LINSTOR supports LVM, LVM-thin, and ZFS, we often get questions about which of the above lower-level storage virtualization technologies we recommend. The answer is, as usual, it depends on your requirements.
If you only look for performance, no matter which features are available, the answer is LVM (traditional). If you want to have snapshots and thin provisioning, we used to recommend LVM-thin before ZFS since, for a long time, ZFS could not keep up with the performance provided by modern NVMe drives.
On the other hand, an example workload that ZFS masters – and where LVM-thin takes minutes(!) – is resizing volumes on a VG/pool with more than 600 logic volumes.
However, since OpenZFS 2.3 (released in 2025), things seemed to have changed in the favor of ZFS. Rene wanted to know for sure and will present the actual performance measurement results of OpenZFS 2.3 improvements in his presentation at the mentioned CSEUG.
Oh, I have one last comment on that topic: do not use ZFS native encryption.
In other recent news, I’m happy to share that Storj has added LINSTOR to their documentation. You can now learn how to integrate LINSTOR with Storj for disaster recovery with step-by-step instructions on the Storj website.
I will close with a recently released blog post, which examines the performance of DRBD in diskless mode. DRBD version 9.0 introduced a feature now used in nearly all LINSTOR integrated clusters: DRBD diskless attachment, otherwise known as DRBD clients.
