[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Why MSDP?
Let me give this a try.
MBGP conveys information about the AS hop path that multicast traffic
should follow. It says that, for example, for us (here at AS 16517) to reach
171.70.89.102 = dhcp-171-70-89-102.cisco.com
multicasttech-2>show ip mbgp 171.70.89.102
BGP routing table entry for 171.68.0.0/14, version 899674
Paths: (2 available, best #2, table NULL)
Flag: 0x208
Advertised to non peer-group peers:
192.168.1.2
145 24 1239 109
204.147.129.89 from 204.147.129.89 (204.147.128.136)
Origin IGP, localpref 100, valid, external
Community: 9503719
1239 109
160.81.38.225 from 160.81.38.225 (144.228.242.54)
Origin IGP, metric 59, localpref 100, valid, external, best
So, for us to get multicast traffic from this IP address, we send the join to AS 1239 (Sprint).
You need MBGP as multicast traffic may need to follow a different route than
unicast traffic. This is the case here :
multicasttech-2>show ip bgp 171.70.89.102
BGP routing table entry for 171.68.0.0/14, version 5801427
Paths: (2 available, best #1, table Default-IP-Routing-Table)
Advertised to non peer-group peers:
192.168.1.2
701 109
157.130.46.53 from 157.130.46.53 (137.39.4.99)
Origin IGP, localpref 100, valid, external, best
1239 1239 1239 109
160.81.38.225 from 160.81.38.225 (144.228.242.54)
Origin IGP, metric 59, localpref 100, valid, external
multicasttech-2>
For unicast, our preferred route involves AS 701 (UUNET).
Now, this is ROUTING. Routing does not convey information about traffic - BGP does not
tell you if you have an email waiting. MBGP likewise does not tell you about sources.
MSDP involves SOURCE DISCOVERY. We have
sources here in AS 16517. We also have an RP. Each of our sources informs our RP of
its existence. Our RP floods this info to our neighboring RP's, including the ones
in Sprint, and they tell the ones in Cisco.com. When 171.70.89.102 wants to
join one of our groups, which it did on Tuesday, it sends an IGMP message which
triggers a PIM-SM join TO THE CISCO.COM RP. This RP knows about us,
and so it sends a join to our source directly.
The biggest advantage to this method is that you rely only on your RP for work. Since
you presumably pay for your RP, this is not a problem.
MSDP uses the MBGP routing info to only select SA messages that come from the right
direction. (We, for example, would only accept CISCO.com SA messages that came through
Sprint.) These Reverse Path Forwarding (RPF) checks for MSDP will likely be relaxed
to make the system more robust - this was discussed a lot at the recent IETF.
SSM shoves the source discovery out of the protocol entirely. You find out sources by other
means (e-mail, webpage,...), and join the group and source directly. MSDP is not needed.
I hate to say it, but this discussion would be more appropriate on the IPMULTICAST
mailing list.
Selina Priestley wrote:
> I am just learning about Multicast, so folks correct me if I am wrong:
>
> I *believe* that MBGP carries the source networks that will be sending
> multicast traffic (there is no tree because BGP is basically fancy
> distance vector). These networks are announced in MBGP whenever the
> sources associated network is up, whether the source is sending traffic
> or not. This is used for reachability info only - how do I forward away
> from this source? or send joins/prunes toward the source?
>
> MSDP is used to send information about specific sources, *when they
> are actually sending traffic*. This is used to start building trees
> in the first place (assuming you have local receivers). Then you use
> the MBGP info to send your join to the source.
>
> Networks are up for days, weeks, maybe even months. Sources are sending
> traffic for hours tops.
>
Not correct, or at least, not always true.
Our sources have been up since October, 2000, and have had continuous multicast state
going out since early November, except for a few hours of downtime due to computer
problems. U Oregon and Broadcast.com, at least, have similar permanent groups.
>
> This may be what you're saying below, but I can't be sure. :)
>
> Selina
>
> >Hello,
> >I think I understand but just correct me if I am wrong.
> >
> >the multicast topology advertised by MBGP is not the Trees but it is just
> >advertising the network reachability of the connected multicast routers. that
> >mean instead of advertising all the routes in a domain , we just advertise the
> >routes between multicast routers and subnetworks.
> >MBGP is not advertising the trees in the domains and that is why we should sen
> >d
> >another kind messages when a source appers. because this source will construct
> > a
> >new multicast tree in the domain using PIM-SM or other protocol.
> >Of course, MBGP like BGP cant support a lot of changes in the topology in a
> >domain and that is why , multicast trees (reachability) should not transported
> >via MBGP.
> >we should found something else (and that is why we are using MSDP)
> >
> >Is that true? (I hope I am right :-) )
> >
> >thanx
> >Ali
> >
> >
Regards
Marshall Eubanks
T.M. Eubanks
Multicast Technologies, Inc
10301 Democracy Lane, Suite 410
Fairfax, Virginia 22030
Phone : 703-293-9624
Fax : 703-293-9609
e-mail : tme@multicasttech.com
http://www.on-the-i.com
Test your network for multicast : http://www.multicasttech.com/mt/