[nsp] relevance of ibgp in private mobile transit backbones?

[Ryan O'Connell]

Please turn off HTML mail when posting to the list. It makes your message
very hard to read.

On Sun, 08 Sep 2002 18:19:18 +0000 padma krishnaswamy <kri234@hotmail.com> wrote:
> <DIV>a question that i hope is not too off topic for this list.</DIV>
> <DIV>The convention (to transmit BGP&nbsp; acquired routes cross
> network)&nbsp;is to use IBGP in transit nets. IBGP is chosen&nbsp;to
> spare the IGP-related complications caused by needing to carry large
> numbers of BGP routes if an IGP were used to do so , and not IBGP. </DIV>
> <DIV>If the scenario is one where </DIV>
> <DIV>1) the number of BGP routes at each border gateway may be much
> lower than for the commercial internet</DIV>
> <DIV>2)the transit network consists of relatively mobile nodes -thus
> making &nbsp;tcp meshes rather tricky to maintain-</DIV>
> <DIV>&nbsp;</DIV>
> <DIV>what scales (total no of routes across all border routers) is it
> worth considering an IGP such as OSPF to carry BGP routes across the
> transit net?</DIV>

I assume by "carry BGP routes across the transit net" you are talking about
non-border routers only and you will still run iBGP between AS border
routers? I don't think it would ever be worth using an IGP rather than BGP
between border routers these days.

How many routes are you talking about here? I woud not imagine any well
designed private network should have more than a couple of thousand routes
in it's BGP routing table - which any competant IGP should be able to
handle as external routes given enough memory. I do not think you would
have a problem unless you have a particularly messy IP allocation strategy
and can't aggregate much.

-- 
      Ryan O'Connell
Mail: ryan@complicity.co.uk
  CV: http://www.complicity.co.uk/ryancv.pdf
      CCIE #8174