RE: [ssm] Dual homed multicast src and SSM

[Beau Williamson <bwilliam@cisco.com>]

But what if the unix box fails??  Wouldn't it be better to have
another box ready to take over the task?  This is anycast sources
and would probably be a better solution.  The idea is that you have
two sources in the network at different locations but using the same IP
address.  The routers connected to these two sources each advertise
routes to the source.  Each half of the network joins the SPT to
nearest source based on the unicast routing information.  (Basic
Anycasting).  Now if the GE link to the first source fails or the
source itself fails then the only remaining route to the source is to
the second source (which is using the same IP address).  Thus, in a
perfect world, all the receivers in the network simple join to (S,G) via
IGMPv3 and SSM. In a not so perfect world, we can accomplish the same
thing with ASM.  We just have to have some RP's to bootstrap the
flow via the Shared Tree and then just let the normal switch to the SPT
do its thing.

If you still want to stick to the way you suggested (and ignore what
happens if the source unix workstation fails) then you would have to
write some code in the unix box to have it bring up the other GE
interface when the first one fails.  Much harder to accomplish than
the Anycast Sources I mentioned above.

Beau
At 10/13/2004 01:55 PM, Field, Brian wrote:

> Sorry for the confusion.
>
> By dual homed I mean a device which originates m-cast content and which
> is
> connected to two routers.  All devices are in the same ASN.  In
> an SSM
> environment, the S in the 
> (S,G) would be an IP associated with this device.
>
> So think of a unix box with two GEs connected to two different routers
> and
> srcing traffic to an SSM channel.  
>
> Since the device has two GEs, I'd like the unix box to be able to
> continue
> to src content to (S,G) even when one of the GE Links fails (or we do
> maint
> on one of the routers, etc.).  Think this means then that the S
> can't be
> associated with a GE physical link IP but more like a loopback like
> IP
> associated with the unix box.  So the question becomes, how do the
> directly
> connected routers Construct the right state on their GE interfaces to
> accept
> the Traffic being src'd by unix box.   
>
> For this to work, does my unix box need to run PIM-SSM?  Or do I
> need to run
> HSRP to 
> have the unix box appear to be an interface on both routers?  Or ...
> ?
>
> Thanks
> Brian
>
>  
>
> > -----Original Message-----
> > From: Beau Williamson
> [mailto:bwilliam@cisco.com]
> > Sent: Wednesday, October 13, 2004 12:21 PM
> > To: Field, Brian; ssm@ietf.org
> > Subject: Re: [ssm] Dual homed multicast src and SSM
> > 
> > Brian,
> > 
> > There is nothing specific to SSM that would prevent the trees 
> > from being setup between one AS and another.  The fact that
> > the source domain is dual-homed only means that there are two 
> > paths form where Joins can arrive (from other domains).  This
> > is of course controlled with normal BGP metrics and is no 
> > different from hosts in other AS's wishing to "send"
> unicast 
> > traffic to the source host.
> > 
> > Beau
> > At 10/12/2004 06:19 PM, Field, Brian wrote:
> > 
> > >Consider a network operating in SSM and a dual homed multicast
> src
> > >device.  Is there a standard way to setup the forwarding
> between the
> > >routers and the src device to get the SSM trees built 
> > properly?  Seems
> > >as if the device needs to eiter participate in routing and 
> > announce the
> > >src IP or run something like HSRP.  Issues with these
> approaches?
> > >Other options?
> > >
> > >Thanks
> > >Brian
> > >
> > >_______________________________________________
> > >ssm mailing list
> > >ssm@ietf.org
> >
> >https://www1.ietf.org/mailman/listinfo/ssm
> > 

_______________________________________________
ssm mailing list
ssm@ietf.org
https://www1.ietf.org/mailman/listinfo/ssm