[c-nsp] mpls and BGP question

[Peter Rathlev]

"Don Hickey" 10/23/07 7:55 PM GMT+2:
<snip>
> On the example above do I need to have a router reflector for this 
> to work? 

You either need to have all your iBGP speakers have all the others 
configured manually (pref. via peer-group) or use a route reflector. 
If you network is small and will stay small for the foreseeably 
future, you could do without the RRs. They introduce a slight delay 
in route propagation, but makes it easier to administer a large group 
of iBGP speakers.

So the short answer is: No, but then you need full mesh BGP 
configuration.

> If so, if PE5 was the RR, would I need to include all the other PE 
> routers as clients? 

Unless you have a very large network, you would have one or two RRs 
(two for redundancy, but one would do fine) and have all others be RR 
clients, peering with the two RRs (or the one RR). (With a very large 
network you might have several levels of RR indirection.)

> I was wanting to use MPLS for our Internet traffic, because in the 
> future their might be other companies that use this network as a 
> transit from one location to another. I would like to keep it 
> seperate from our traffic. 

You're probably referring to "MPLS VPN", which can give you traffic 
seperation among other things. You can do transit fine without MPLS 
VPN; if you would only transport internet routable addresses, there 
shouldn't be any need for MPLS VPN. If you would transport private 
addresses you would probably need something like MPLS VPN.

> I have no problems reading if someone were to have a nice link to 
> send me. I have been working on this for too long and would like to 
> knwo if I am on the right track or if there is a better way to 
> implement a solution. 

You can read a little about BGP peer groups and RR here:
http://www.tinyurl.dk/1988

Look for the "Controlling the Flow of BGP Updates" chapter.

Regards,
Peter Rathlev