Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 26 Jan 2004 22:52:08 -0600
From:      Tillman Hodgson <tillman@seekingfire.com>
To:        current@freebsd.org
Cc:        performance@freebsd.org
Subject:   Re: Old SUN NFS performance papers.
Message-ID:  <20040127045208.GC83558@seekingfire.com>
In-Reply-To: <Pine.NEB.3.96L.1040124210942.31483E-100000@fledge.watson.org>
References:  <003c01c3de8d$d569edb0$471b3dd4@dual> <Pine.NEB.3.96L.1040124210942.31483E-100000@fledge.watson.org>

next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, Jan 24, 2004 at 09:14:51PM -0500, Robert Watson wrote:
> I haven't done much benchmarking on NFS lately, but something worth
> remembering is that people have spent a lot of time researching and
> optimizing TCP for a variety of connection types, whereas the NFS code
> basically has a static implementation of RPC backoff and flow control that
> hasn't evolved much.  TCP is aware of things like the pathwise-mtu to the
> server and adapts, whereas UDP just loses packets due to fragmentation,
> especially if you are using larger block sizes.  Please do post your
> discoveries on performance@, and perhaps we could build an NFS performance
> tuning section in the FreeBSD Handbook (or if there's not that much
> content, add it to the FAQ)?

I once spent a great deal of time with bonnie++, gnuplot and LaTeX
generating a nice 10-page or so document on NFS tuning between various
systems (the server was always a FreeBSD 4-STABLE box of the 4.7 vintage
with several vinum mirrors of various speeds).

Naturally, I then lost the document when my home drive disintegated a
few months later. i was /just/ in the process of moving my home dir to
the file server where it would have multiple levels of backup ...

Anyway, I'd be willing to do some more testing and writes up the results
for the Handbook (when I get back from a much-needed vacation) if I
could obtain several collaborators with a variety of hardware (NFS
testing takes /much/ time to cover all the variants and hardware
possibilities) and a good idea from someone (such as yourself) on what
sort of testing would be seen as valuable.

For instance, my primary concern is remote Maildir speed and backup up
several servers onto a backup host (where it can spooled out to tape
via some scripts). So my access patterns are fairly specific, and likely
not typical. Guidance on what a typical NFS user looks like would be
precious knowledge ;-)

-T


-- 
Page 38: Be sure that, in the excitement of creating a totally rad
password, you resist the temptation to tell someone just to show off how
smart you are.
	- Harley Hahn, _The Unix Companion_



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20040127045208.GC83558>