Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 13 Feb 2007 13:04:06 -0500 (EST)
From:      Rick Macklem <rmacklem@uoguelph.ca>
To:        drumslayer2@yahoo.com
Cc:        freebsd-fs@freebsd.org, Chris Haulmark <chris@sigd.net>, Eric Anderson <anderson@freebsd.org>
Subject:   Re: UFS2 with SAN
Message-ID:  <Pine.GSO.4.63.0702131247150.8743@muncher>
In-Reply-To: <a969fbd10702130800i5f181b92g2eb8157b15330f10@mail.gmail.com>
References:  <45CD6FF5.8070007@freebsd.org> <20070213075627.63126.qmail@web34502.mail.mud.yahoo.com> <a969fbd10702130800i5f181b92g2eb8157b15330f10@mail.gmail.com>

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

> On Tue, 13 Feb 2007, Jeff Mohler wrote:
> For certain workloads, linux is far better tuned for NFS than Fbsd..netapp
> as a server or not.
> On 2/12/07, Nicole Harrington <drumslayer2@yahoo.com> wrote:
>>
>> 
>> I have a set of servers NFS mounted to a Netapp and
>> after hurs of tuning with netapp's help. (after
>> getting through the idiots adking what FreeBSd was)
>> I got very low performance. I was of course then told
>> by Netapp to switch to Linux for better NFS support.
>> 
>> I would love for any help with tuning this further,
>> but I cannot say that FreeBSD with Netapp NFS will be
>> great. Of course, I have not been able to test if
>> indeed Linux would be any better.
>> I will say however that I have a large number of
>> small files which tends to not do well with NFS.
>> 
>> 
>> Nicole

I don't know if any of this will help, but...
My experience is that abismal NFS perf is usually
caused by one of:
- Transport/Network problems: If the network interconnect
   introduces delays or drops packets, you've got to fix it.
   This includes network device drivers that aren't up to it.
   If you use NFS over UDP, one thing to check for is the # of
 	"fragments dropped due to timeout" in the IP section
   of "netstat -s". If there are any of those happenning, you
   need to fix your network, or at least switch to TCP or
   reduce your read/write data size. (Make rsize, wsize as
   large as you can, given what your network can handle.)

- Byte range locking (the NLM protocol, not really NFS). It's
   a dog in mho. All you can do is avoid apps that do byte range
   locking. (NFSv4 has good byte range locking support, but the
   client in FreeBSD isn't useful for production work yet, from
   what I've seem.)

Beyond that, I'm not sure there is much you can do, rick




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.GSO.4.63.0702131247150.8743>