X-MimeOLE: Produced By Microsoft Exchange V6.5
Received: by onstor-exch02.onstor.net 
	id <01C7837A.352A2500@onstor-exch02.onstor.net>; Fri, 20 Apr 2007 11:32:10 -0700
MIME-Version: 1.0
Content-Type: text/plain;
	charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
Content-class: urn:content-classes:message
Subject: RE: Poor NFS performance.  Dell Ethernet Switches.
Date: Fri, 20 Apr 2007 11:32:10 -0700
Message-ID: <BB375AF679D4A34E9CA8DFA650E2B04E03033EDF@onstor-exch02.onstor.net>
In-Reply-To: <BB375AF679D4A34E9CA8DFA650E2B04E0351C1EA@onstor-exch02.onstor.net>
X-MS-Has-Attach: 
X-MS-TNEF-Correlator: 
Thread-Topic: Poor NFS performance.  Dell Ethernet Switches.
Thread-Index: AceC+fKJyXNbIqPBSQ2eWS3xFi245AAAiIg2AA1ycyAACF75IAAItu1Q
From: "Jeseem S" <jeseem@onstor.com>
To: "Steffen Thuemmel" <steffen.thuemmel@onstor.com>,
	"Tim O'Callaghan" <tim.ocallaghan@onstor.com>,
	"John Culp" <john.culp@onstor.com>
Cc: "dl-cstech" <dl-cstech@onstor.com>,
	"John Rogers" <john.rogers@onstor.com>

Is the dell switch configured as L3 also or plain L2 trunk.

Can you pull down (physically remove cable) on one port in the trunk and
see if the performance improves.

Do you see any packet retransmissions in filer?
Can you check the stats =20
In cli : stats trxr tcp
 And txrx console : tpl show stats

The mode must be set to 'on' only ( since we cannot talk PAgp or LACP)
Can you send the output of show ether-channel (or port group) and show
spanning-tree and show trunks

-Jeseem


-----Original Message-----
From: Steffen Thuemmel=20
Sent: Friday, April 20, 2007 7:00 AM
To: Tim O'Callaghan; John Culp
Cc: dl-cstech; John Rogers
Subject: RE: Poor NFS performance. Dell Ethernet Switches.

John,

I'm running Dell switches in my lab. I never had performance problems
caused by them. Also the trunking just works fine.
A similar scenario like you described is experienced by DIR.BG. Up to
now no root cause had been identified. See case 4574.

St.

Steffen Thuemmel=20
OnStor

mobil. +49 173 673 3434
mail. steffen.thuemmel@onstor.com


-----Original Message-----
From: Tim O'Callaghan=20
Sent: Freitag, 20. April 2007 11:59
To: John Culp
Cc: dl-cstech; John Rogers
Subject: RE: Poor NFS performance. Dell Ethernet Switches.

John,

From my experience of the Dell switches, especially the later ones - the
command is almost identical to the Cisco "channel-group" rather than
"port group"(it is the same code base I believe)

channel-group <# group id number> mode on

Mode "on" forces channel group membership with no LACP

Regards

Tim



-----Original Message-----
From: John Rogers=20
Sent: 20 April 2007 04:29
To: John Culp; dl-cstech
Subject: Re: Poor NFS performance. Dell Ethernet Switches.

Lacp is not supposed to work. I would suspect that any lacp config on a
dell switch would cause packet loss. That will kill performance. =20

I have several dell's in the lab. I'll need to review the aggregation
setup. But its really similar to cisco on the command line use ether
channel groups and set them to mode on.=20

John

-----Original Message-----
From: John Culp
To: dl-cstech
Sent: Thu Apr 19 20:14:03 2007
Subject: Poor NFS performance.  Dell Ethernet Switches.

IHAC that is seeing really poor NFS performance.  They are using a
Xyratech array that I know is capable of good performance (not the new
one that does not work).  I am curious if anyone has a customer using
Dell Ethernet switches?  The LACP never works, Trunking is difficult and
it just seems to be the slow down.
=20
We are doing a dd to the filer via a NFS client and Performance is 50 MB
for 4 seconds and then drops to 10MB.
=20
Any thoughts?
=20
John Culp=20
Sr System Engineer
john.culp@onstor.com <mailto:john.culp@onstor.com>=20
office: 940-239-7489
cell: 972-523-4287
ONStor, Inc.
NAS for Enterprise
3311 Chalkstone Cove| Corinth, TX 76208
main: 940-239-7489 | fax: 940-239-7527=20
=20
