[c-nsp] ISP / MPLS "POP" design

[CiscoNSP List]

Thanks very much Mark - Appreciate your input.


> From: mark.tinka at seacom.mu
> To: cisconsp_list at hotmail.com
> Subject: Re: [c-nsp] ISP / MPLS "POP" design
> Date: Thu, 31 Oct 2013 04:06:27 +0200
> CC: cisco-nsp at puck.nether.net
> 
> On Wednesday, October 30, 2013 11:35:59 PM CiscoNSP List 
> wrote:
> 
> > Thanks Mark.
> > 
> > So to clarify - If I run 2 (7201's) as RR's, they would
> > take the full tables from the IPTransit 7200's(POPA),
> > plus all customer global IP's, plus all VPNv4
> > routes(From POPA+B+C)?
> 
> That's right.
> 
> The 7201's CPU is very capable. It tops out at 2GB of RAM, 
> and should be able to handle your current deployment just 
> fine.
> 
> > If that's the case - Do you filter what routes the RR's
> > advertise to RR clients?  i.e. POPA has the 2 7200's
> > with IPTransit full table, do the RR's advertise the
> > full table to the 7200's at POPB + C?
> 
> That's a design decision.
> 
> Some operators don't filter in iBGP, ensuring every router 
> has, pretty much, the same view of the state of BGP in the 
> core.
> 
> Other operators, like myself, implement network-wide routing 
> policy in iBGP, which is easiest done in the route 
> reflectors, as that is how different routers performing 
> different functions learn (which) routes (they should be 
> receiving).
> 
> If you're not sure what to do, go simple and evolve the 
> configuration as your network does so.
> 
> Cheers,
> 
> Mark.