AF:
NF:0
PS:10
SRH:1
SFN:
DSR:
MID:<20070914110617.626ad52f@ripper.onstor.net>
CFG:
PT:0
S:andy.sharp@onstor.com
RQ:
SSV:onstor-exch02.onstor.net
NSV:
SSH:
R:<eric.barrett@onstor.com>,<john.rogers@onstor.com>,<brian.montero@onstor.com>,<fay.chong@onstor.com>,<paul.hammer@onstor.com>,<brian.baker@onstor.com>,<maxim.kozlovsky@onstor.com>,<brian.deforest@onstor.com>
MAID:1
X-Sylpheed-Privacy-System:
X-Sylpheed-Sign:0
SCF:#mh/Mailbox/sent
RMID:#imap/andys@onstor.net@onstor-exch02.onstor.net/INBOX	0	BB375AF679D4A34E9CA8DFA650E2B04E0587B1D6@onstor-exch02.onstor.net
X-Sylpheed-End-Special-Headers: 1
Date: Fri, 14 Sep 2007 11:06:59 -0700
From: Andrew Sharp <andy.sharp@onstor.com>
To: "Eric Barrett" <eric.barrett@onstor.com>
Cc: "John Rogers" <john.rogers@onstor.com>, "Brian Montero"
 <brian.montero@onstor.com>, "Fay Chong" <fay.chong@onstor.com>, "Paul
 Hammer" <paul.hammer@onstor.com>, "Brian Baker" <brian.baker@onstor.com>,
 "Maxim Kozlovsky" <maxim.kozlovsky@onstor.com>, Brian DeForest
 <brian.deforest@onstor.com>
Subject: Re: Free BSD for Yahoo test
Message-ID: <20070914110659.15d9dccc@ripper.onstor.net>
In-Reply-To: <BB375AF679D4A34E9CA8DFA650E2B04E0587B1D6@onstor-exch02.onstor.net>
References: <BB375AF679D4A34E9CA8DFA650E2B04E0346A97B@onstor-exch02.onstor.net>
	<BB375AF679D4A34E9CA8DFA650E2B04E0587B1D6@onstor-exch02.onstor.net>
Organization: Onstor
X-Mailer: Sylpheed-Claws 2.6.0 (GTK+ 2.8.20; x86_64-pc-linux-gnu)
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit

On Fri, 14 Sep 2007 09:42:44 -0700 "Eric Barrett"
<eric.barrett@onstor.com> wrote:

> "Something is majorly wrong with NFS."

A while ago, Max did some poking around using ripper (2.6.22.3)
and figured out at least a narrowing of the circumstances for this
problem.  Like it only happens when copying from and to a filer, as
opposed to from only or to only.

This shouldn't be that hard a problem for someone in the protocols
group to figure out, that's what they do.

> In the Linux world, distribution maintainers like RedHat and Debian
> typically add extra patches to their kernels to make up for flaws in

Dude.  I thought I taught you better than that.  Sheez, teacher gets an
'F'.

Please don't group this week's Redhat intern charged with putting
together a kernel with a Debian kernel package maintainer.  Also try to
remember that a distro isn't a kernel and vice versa.  I run Debian but
I don't run one of their kernels.  So this isn't a distro-centric issue.

> A good example of how badly kernels can break -- we installed stock
> Linux 2.6.22 on beast (the data warehouse system) and dd performance
> dropped from 110MB/sec to 2.  Yes, two.

Doesn't imply a breakage in the kernel.  More like a breakage in our
filer.  Although maybe your implication that Trondie is slipping in
some NFS code that would give competitors a headache is always possible.

> Since Yahoo is the only company of any size who uses FreeBSD, the
> situation is worse.  There aren't a million splinters of FreeBSD out
> there you can try until you find one that works.  And the only company
> in the industry with experience tuning FreeBSD is...Yahoo.  (But
> they're very good at it.)

Companies don't have experience, people do.  We can certainly set up a
FreeBSD box to test with.

> I used to work with the Yahoo support team at NetApp, and I can tell
> you NetApp did very little performance troubleshooting.  They sold on
> features and price, and let Yahoo figure the performance stuff out.
> It may not be a viable approach for us, but that's what happened.

I will point out that Yahoo is dumping FreeBSD and moving to Linux.
And there is 5 gazillion Linux machines out there.  Our product has to
work with Linux and Windows w/o having to custom tune those boxes.

Cheers,

a
