[Btrfs-users] btrfs benchmarks

[Chris Mason]

On Thursday 24 January 2008, Ales Blaha wrote:
> Hi,
>
> I`ve find about BtrFS just this week, so I`ve not tested it so far. I`ll do
> it as soon as I got a spare disk to experiment with. But, I`ve two
> questions regarding BtrFS. First, do you plan inclusion of BtrFS into
> mainline kernel and if so, when do you expect this to happen? 

I plan on submitting btrfs for inclusion into the mainline kernel when the 
disk format is finalized and I believe the code is stable.  A rough timeline 
of the pending disk format changes can be found here:

http://oss.oracle.com/projects/btrfs/dist/documentation/todo.html

> Second, I 
> would like to see some more benchmarks of BtrFS, so far you provided
> comparison to Ext3 and XFS, which is good, but, as far as I understand one
> of the main BtrFS targets is performance on small files, especially when
> there are large number of them. Ext3 is a general purpose FS, and as such
> it deals with small files rather inefficiently. XFS handles large number of
> files pretty well, however it has been optimised for opposite type of
> workload, that is for streaming very large files. Hence it would be more
> telling to compare performance of BtrFS with a file systems that targets
> similar workloads. Most prominent example is perhaps ReiserFS, especially
> its 4th version.

More benchmarking will be done as after the next round of disk format changes 
to support multiple devices.  I don't plan on testing against reiser4, but 
I'm happy to provide links to other pages that benchmark across more 
filesystems.  This is really just a factor of the time available, 
benchmarking is time consuming.

Also, the goal of btrfs is not to beat every filesystem in every benchmark.  
The goal is to keep up with the linux filesystems that already perform very 
well and to provide a large set of new features at the same time.

-chris