Re: XFS write speed drop

[Nathan Scott <nathans@sgi.com>]

On Wed, May 24, 2006 at 08:22:19AM +1000, Nathan Scott wrote:
> On Tue, May 23, 2006 at 03:23:31PM +0200, Jan Engelhardt wrote:
> > >> CASE 1: Copying from one disk to another
> > >> ========================================
> > >> Copying a compiled 2.6.17-rc4 tree; 306907 KB in 28566 files in 2090
> > >> directories.
> > >
> > >OK, we can call this a metadata intensive workload - lots of small
> > >files, lots of creates.  Barriers will hurt the most here, as we'd
> > >already have been log I/O bound most likely, and I'd expect barriers
> > >to only slow that further.
> > >
> > Yes and the most important thing is that someone made -o barrier the 
> > default and did not notice. Someone else? :-D
> 
> Not sure what you're trying to say here.  Yes, barriers are on
> by default now if the hardware supports them, yes, they will
> slow things down relative to write-cache-without-barriers, and
> yes we all knew that ... its not the case that someone "did not
> notice" or forgot about something.  There is no doubt that this
> is the right thing to be doing by default - there's no way that
> I can tell from inside XFS in the kernel that you have a UPS. ;)

Oh, I realised I've slightly misread your mail, you said...

| I do not actually need barriers (or an UPS, to poke on another thread),
| because power failures are rather rare in Germany.

Hmm, even harder for us to detect at runtime, in the kernel,
that you're in Germany. :)

Power failures aren't the only thing to cause crashes, however.

cheers.

-- 
Nathan