From: Rick Burts (burts@xxxxxxxxxxxxxx)
Date: Thu Nov 01 2001 - 22:41:23 GMT-3
Jim
I think your logic is quite correct (and I have done DLSW in a production
environment). Where there is a single interface involved with DLSW it is
more effective and efficient to peer to the single interface and when
there are multiple interfaces it is best to peer to a loopback.
This will be my last post from this address as Mentor has shut down. I
will subscribe to the list with a new address as my situation gets
straightened out.
Rick
On Thu, 1 Nov 2001, Jim Brown wrote:
> Imagine this.... you are peering between sites 'A' and 'B' using loopbacks
> across a WAN link. The Ethernet port at site 'A' fails. Site 'B' will
> continue to send data across the WAN link. From the perspective of site 'B'
> everything is OK because the DLSW connection is maintained between
> loopbacks.
>
> The DLSW connection is completely unaware of the fact that the data has
> nowhere to go once it arrives at site 'A'. I don't believe DLSW monitors the
> availably of interfaces attached to the ring group/bridge group.
>
> Using the same scenario except we peer between the Ethernet ports. If one of
> the Ethernet ports fail, the DLSW connection is torn down conserving our
> precious WAN bandwidth.
>
> This is my reasoning why it is better to peer between physical interfaces in
> certain circumstances. All bets are off with two physical interfaces on the
> same router participating in DLSW. You could potentially tear down the
> connection when only one of the interfaces fails and leave the other segment
> on the same router out in the cold.
>
> I'm only speaking of theory. I might be way off my rocker. I've never used
> DLSW in a production environment. So clarification of my ramblings are
> welcomed. If a DLSW guru wants to straighten me out I'm all ears.
>
> -----Original Message-----
> From: Jason Sinclair [mailto:sinclairj@powertel.com.au]
> Sent: Thursday, November 01, 2001 6:27 PM
> To: 'fwells12'; Jim Brown; ccielab@groupstudy.com
> Subject: RE: DLSW peer question
>
>
> However,
>
> Based on the logic of single interfaces, if the single interface were to
> fail (and they were directly connected!) the a side wouldn't send anyway as
> layer 2 would have failed?
>
> Regards,
>
> Jason Sinclair
> Network Support Manager
> POWERTEL Limited
> Level 11, 55 Clarence Street, SYDNEY
> Phone: 61-2-8264-3820
> Fax: 61-2-9279-2604
> Mobile: 0416 105 858
> jasons@powertel.net.au
>
> -----Original Message-----
> From: fwells12 [mailto:fwells12@hotmail.com]
> Sent: Friday, 2 November 2001 10:02
> To: Jim Brown; ccielab@groupstudy.com
> Subject: Re: DLSW peer question
>
> Good point.
>
> ----- Original Message -----
> From: "Jim Brown" <Jim.Brown@CaseLogic.com>
> To: "'Geir Jensen'" <geir@hfk.vgs.no>; "fwells12"
> <fwells12@hotmail.com>;
> <ccielab@groupstudy.com>
> Sent: Thursday, November 01, 2001 1:45 PM
> Subject: RE: DLSW peer question
>
>
> > In a production environment would it not make more sense
> to tie the local
> > peer to a single physical interface if there is only one
> interface
> > participating in DLSW?
> >
> > If it is tied to the physical interface, when it goes down
> then so does
> the
> > DLSW connection.
> >
> > If it is tied to a loopback and the physical interface is
> down, DLSW
> traffic
> > will travel to the remote end only to be dropped.
> >
> > It seems to me tying it to the physical interface would
> conserve bandwidth
> > during interface failures.
> >
> >
> > -----Original Message-----
> > From: Geir Jensen [mailto:geir@hfk.vgs.no]
> > Sent: Thursday, November 01, 2001 2:16 PM
> > To: fwells12; ccielab@groupstudy.com
> > Subject: RE: DLSW peer question
> >
> >
> > I always use the loopback, it's definately more stable
> than the
> > alternatives. Geir Jensen
> >
> > -----Original Message-----
> > From: fwells12
> > Sent: Thu 11/1/2001 9:34 PM
> > To: ccielab@groupstudy.com
> > Cc:
> > Subject: DLSW peer question
> >
> >
> >
> > OK, you have a router which has more than one LAN
> interface it needs
> > DLSW
> > traffic forwarded from, let's say e0 and t0. It also has
> a loopback
> > int with
> > ip ad 172.16.10.1/24. The ip's of the e0 and t0
> interfaces are
> > 172.16.20.1/24
> > and 172.16.30.1/24 respectively. What ip address is the
> best
> > practice to use
> > for your dlsw local-peer peer-id statement?
> >
> > Cheers
>
This archive was generated by hypermail 2.1.4 : Fri Jun 21 2002 - 06:45:01 GMT-3