[Novalug] btrfs vs the world ...

[Gregory Maxwell]

On Tue, Dec 15, 2009 at 8:12 AM, Varol Okan
<varokan at movingsatellites.com> wrote:
> I was planing in playing with btrfs, to see performance but after the
> Phoronix test results, I guess this would be wasted time ( wrt
> performance anyways ).

"Phoronix test results" is usually synonymous with "worthless", sadly.

They often use daft methodology, and insufficiently report what they tested so
that no one can reproduce it, or even have a fighting chance of fixing any real
problems that exist.

One challenge in FS testing is that different FSes provide different features
by default.  Journaled FSes often provide greater safety... BRTFS provides
checksum protection by default, etc. So some knob tweaking is required
to provide an apples to apples comparison. It looks pretty obvious that
they didn't do this.

I have BTRFS on some of my development boxes. Between the COW files and the
compression it saves save a TON of disk-space from many duplicate
development/testing source code copies. But really, it's still early and
immature. Don't expect it to not have performance warts, or expect it to not
eat your data.