Daniel Nashed's Blog

Anyone using ZFS on Linux for Domino?

ZFS (https://openzfs.org) is a quite interesting file system. I have looked into it before but never had the strong requirement to use it.
Beside the license challenges and that it isn't part of the major distributions, I really really like what I see on first look.

For Domino Backup we have another good reason to look into ZFS.
Beside BTRFS this is the only Linux file system with out of the box snapshot support.
For ext4 you can eventually get it working with LVM snapshots. But this doesn't look like a scalable good to maintain approach.

I have a customer running on Proxmox virtualization with ZFS for a long time. And they are quite happy.
For them it was a logical choice because Proxmox is leveraging ZFS in their platform.

ZFS makes a lot of sense an different settings make sense for different environments and data loads.
There are mainly 3 aspects I am interested in:

• Snapshots
  
  This is my current main motivation. With Domino Backup having a way to snapshot the file system would be a very convenient way to backup and restore.
  What ZFS offers is exactly what we need.
  
  - command-line integration for creating a snapshot
  - tagging snapshots with a reference to find them
  - mounting snapshots for restore
  
• Deduplication
  
  When it comes to Domino live data, deduplication can save you up to 30% of space.
  But the down side would be reduced performance and higher RAM requirements.
  So probably deduplication is not a good idea for live data.
  
• Compression
  
  Similar to deduplication this would save some disk space but also reduces the performance.
  I have tested with my own 10 GB mail file that compression can provided the same 30% reduction and lead to zero deduplication
  

But there are other advantages as well

• On-line Defragmentation
• Pooled easy to use flexible storage
• Easy to extend
• Support for additional data sets for resilience

Deduplication & Compression not suitable for NSF & Co - But as a backup target!

Deduplication and Compression don't sound like a good match for Domino NSF, translog, NIF or FT data.
But they sound like a very good option for DAOS files when hosted for multiple servers -- for example one server hosting them over NFS.

If using ZFS for a remote storage pool to backup your Domino data (based on snapshots) sounds like a great solution offering high level of deduplication.
I have tested with databases modified over time (without a copy style compact of course). And the data deduplication rate was very high.


Snapshots

My current take is Domino Backup optimization.

ZFS offers a very easy to use and flexible interface to create snapshots.
And they can be mounted as a volume on the fly.

So they can be either used as backup or to provide a consistent source for backup with Domino V12 backup in snapshot mode.


Your feedback

I am currently looking into it. But I would really like to see if anyone out there is using ZFS for Domino production work-loads already.
And specially what experience you have with features like deduplication and compression.

• Comments [3]