Does anyone have some methods for doing diagnostic / performance
profiling while using NFS over TCP?
Thanks in advance,
Matthew Sacks
matthewsacks.com
On Tue, Jul 15, 2008 at 1:18 PM, ntwrkd <[email protected]> wrote:
> Does anyone have some methods for doing diagnostic / performance
> profiling while using NFS over TCP?
There's nfsstat. Do you want something more than that?
--
Andrew Bell
[email protected]
Yes
------Original Message------
From: Andrew Bell
Sender:
To: [email protected]
Sent: Jul 15, 2008 11:31 AM
Subject: Re: NFS performance / diagnostic profiling with TCP
On Tue, Jul 15, 2008 at 1:18 PM, ntwrkd <[email protected]> wrote:
> Does anyone have some methods for doing diagnostic / performance
> profiling while using NFS over TCP?
There's nfsstat. Do you want something more than that?
--
Andrew Bell
[email protected]
You could use a tool like iozone, bonnie or nfsreplay to
generate a workload against the server and collect profiling samples
using oprofile. I ran a bunch of tests some time back using nfsreplay,
see the results on the page below as a sample of the kind of data you
could generate using oprofile:
http://www.gelato.unsw.edu.au/IA64wiki/NFSServerPerformance#head-b5a633e0fccdde424b0395181f90efa08d56d048
Shehjar
ntwrkd wrote:
> Does anyone have some methods for doing diagnostic / performance
> profiling while using NFS over TCP?
>
> Thanks in advance, Matthew Sacks matthewsacks.com -- To unsubscribe
> from this list: send the line "unsubscribe linux-nfs" in the body
> of a message to [email protected] More majordomo info at
> http://vger.kernel.org/majordomo-info.html
ntwrkd wrote:
> Does anyone have some methods for doing diagnostic / performance
> profiling while using NFS over TCP?
>
Using tcpdump with tcptrace is really handy.
You do something like:
tcpdump -s 100 -w /tmp/tcpdump.out host <hostname>
tcptrace -Sl /tmp/tcpdump.out
xplot /tmp/a2b_tsg.xpl
One very important thing to remember, if you are interested in how the
tcp window is scaling, is to start tcpdump before you mount the server.
The tcp window size depends on a scaling factor that it gets when the
tcp connection is created (which is at mount time).
Dean
> Thanks in advance,
> Matthew Sacks
> matthewsacks.com
> --
> To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>