ext4: Still Not There Yet – And Not Worth It Anyway
When reading Ext4 Within Striking Distance of XFS the other day, I could not help virtually patting Matthias Eckermann, who owns the storage strategy for SUSE Linux Enterprise on my team, and our file system engineers led by Jeff Mahoney on their backs.
This is a nice confirmation of us shipping XFS as a supported file system for four major releases, from SUSE Linux Enterprise Server 8 to version 11 today.
And as Matthias has analyzed in detail, Btrfs which debuted in SUSE Linux Enterprise Server 11 SP1 is the future of local file systems in Linux and beyond. Btrfs shall move to fully supported state with SP2 and is going to beat ext4 (and Oracle’s ZFS, by the way) hands down in terms of feature/function and be second to none in terms of performance.
So, it’s ext3 and XFS today and Btrfs tomorrow. There is no need ext4 really serves.
In fact that other Enterprise Linux vendor just confirmed this yesterday, announcing support for XFS as an add-on for US$199 per socket-pair per year. Yes, that’s right, additional costs for what SUSE Linux Enterprise has been providing for four major releases as part of the basic offering.
Thanks Matthias and Jeff, good technology vision and execution. I’m glad to have you guys around. 😉
d’Ehre, and sorry, Ted!
PS: Looking into the benchmark that triggered this blog, it turns out the ext4 results approaching, but still not reaching, XFS were without a journal. Which is a big, fat invitation for data corruption (and a slooow fsck run to try recover after the fact). I feel strongly that we should not push file systems that force such a choice between performance and data integrity when there are better solutions such as XFS and Btrfs.