From: John (veryconnected@xxxxxxxxx)
Date: Fri Sep 22 2000 - 09:44:27 GMT-3
I'm not sure what your quesion is? I have some
history with multicasting at a large bank we ran pim
sparse mode with auto rp and then static rp. we found
we had less problems with the static rp defined. just
create loopbacks on both routers and weight one with a
hi bandwith. this will keep the RP constant. If
theres a failure the other will pick up. It works!
--- Mark Lewis <markl11@hotmail.com> wrote:
>
>
> Right, big question!
>
> Rendezvous Point are a common distribution point for
> multicast traffic when
> using a sparse mode multicast routing protocol (eg.
> PIM SM/CBT (although
> with CBT it has a different name).
>
> With multicast there are three important levels of
> interaction:
>
> 1.Interaction between hosts on subnets and 'local'
> routers (registration not
> of clients (only the last reporter is registered per
> group),but of multicast
> groups (traffic) that hosts on that segment
> require). This process is
> handled by IGMP.
>
> 2.There's interaction between switches and local
> routers via CGMP or IGMP
> snooping (registration of multicast clients on the
> switches so that
> multicast traffic is not just flooded (default
> behavior). This (CGMP/IGMP
> snooping) is optional, but recommended.
>
> 3.Interaction between routers.
>
> a. Dense mode (eg. DVMRP/PIM DM/MOSPF) : traffic
> is flooded throughout the
> network initially, until 'downstream' routers tell
> upstream ones that they
> don't want the traffic (pruning). Routers use RPF
> checks to switch over to
> SPT (shortest path tree) back to source.
>
> b. Sparse mode: traffic here is not flooded.
> 'Downstream' routers here
> request traffic from the RP intially and then switch
> over to SPT (shortest
> path back to 'real' source (ie. multicast server)
> when they know where it is
> (they find out after they get initial packets from
> the RP by looking at the
> source address of the packets).
>
> It's best to have a few RP candidates on the network
> (use auto rp instead of
> manual rp config on large net.s) for redundancy. Can
> be in core of net or
> close to multcast servers.
>
> Hope that helps,
>
> Mark
>
>
>
> >From: "Eddie Parra" <eparra@telocity.com>
> >Reply-To: "Eddie Parra" <eparra@telocity.com>
> >To: "CCIE Group Study" <ccielab@groupstudy.com>
> >Subject: Multicasting Questions...
> >Date: Wed, 20 Sep 2000 10:23:34 -0400
> >
> >Can anyone answer a few multicasting questions. Am
> I correct in saying
> >that
> >the RP handles registration for all multicast
> clients? My other question
> >is, where do you place the RP's? Do I need a RP on
> every segment? Every
> >VLAN? Could someone explain this a little better.
> Thanks...
> >
> >-Eddie
> >
> >
>
This archive was generated by hypermail 2.1.4 : Thu Jun 13 2002 - 08:25:00 GMT-3