Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 18 Mar 2003 10:45:40 -0500 (EST)
From:      Daniel Ellard <ellard@eecs.harvard.edu>
To:        John <jwd@bsdwins.com>
Cc:        Hackers List <freebsd-hackers@freebsd.org>, dwmalone@maths.tcd.ie
Subject:   Re: NFS performance tuning
Message-ID:  <Pine.BSF.4.51.0303181032010.45566@bowser.eecs.harvard.edu>
In-Reply-To: <20030318020201.GA2329@BSDWins.Com>
References:  <20030318020201.GA2329@BSDWins.Com>

next in thread | previous in thread | raw e-mail | index | archive | help

The tweaks I've been working on are for read performance.  Based on
what you sent, I don't think read performance is your problem at all
(although they might help you anyway, in the long run).  So, my advice
might be no more useful than line noise, but here goes:

1.  You've got a nfsd taking 48% of an athlon 2200?  Wow.

	If you can profile it to get a look at where it's spending the
	time, that would be incredibly interesting.  I've never seen
	an nfsd do that.

2.  You've got 5 network interfaces, and 2 of them are gigabit, but
	you're only running 8 nfsds.  By rule of thumb, I would add
	more.  This doesn't solve your mystery, however, because it's
	the master that's eating all the CPU.

	(Why do the gigabit interfaces see an order of magnitude less
	traffic than the 100Mb/s?)

3.  Are you using TCP or UDP?  Any kind of flow control on the switch?

I'm interested to hear what other people have to say.

-Dan


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-hackers" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.BSF.4.51.0303181032010.45566>