Article on "next-gen" filesystems (ZFS)

[krikboh]

(updated) Article on next generation file systems like ZFS and btrfs https://www.ixsystems.com/blog/open-zfs-vs-btrfs/

 

[warri]

Interesting read, thanks for sharing!

 

[joeschmuck]

Read it as well. Interesting and looks like we have the better (not experimental) file system right now.

 

[scurrier]

Great article. Thanks for posting it! Love Ars Technica.

I might be becoming a file system nerd.

 

[Knowltey]

Their bit on compression seems a little off. Says that in ZFS you can only enable compression across the entire filsystem. But just a couple of minutes ago I turned off compression for just only one of my datasets.

 

[warri]

They are mentioning "filesystem level", but I'm pretty sure that they actually include datasets in this definition.
Their main point is that Btrfs allows compression on the file-level, or on certain file types - which ZFS cannot do.

 

[titan_rw]

Their mentioning zfs only supports different compression options at a filesystem level caught me too. As warri also pointed out, I realized that they are probably talking about datasets. A dataset is basically a completely different filesystem. So what they state is technically true, although a bit misleading.

I'll concede that zfs doesn't allow you to set compression based on file extensions and such. But then again, I don't see a huge need for that. Datasets allow you different locations to store file-extensions that may or may not compress. Most of my files are already compressed media, so I leave compression off completely. Backup datasets that I know are going to receive uncompressed datastreams, I set to gzip-max. There's no need to do that one a file-extension basis. Other backups that are already compressed by the backup application, I leave compression off.

 

[warri]

I'll concede that zfs doesn't allow you to set compression based on file extensions and such. But then again, I don't see a huge need for that. Datasets allow you different locations to store file-extensions that may or may not compress. Most of my files are already compressed media, so I leave compression off completely. Backup datasets that I know are going to receive uncompressed datastreams, I set to gzip-max. There's no need to do that one a file-extension basis. Other backups that are already compressed by the backup application, I leave compression off.

Also with the introduction of lz4 compression you can basically leave compression on everywhere without a performance penalty, because it aborts early on incompressible data - which is the new default behaviour in the upcoming FreeNAS 9.2.1, see feature #3872.