[172415] in North American Network Operators' Group
Re: Help with Confederation-RR-MPBGP
daemon@ATHENA.MIT.EDU (Philip Lavine)
Wed Jun 18 12:33:46 2014
X-Original-To: nanog@nanog.org
Date: Wed, 18 Jun 2014 09:31:38 -0700
From: Philip Lavine <source_route@yahoo.com>
To: "mh@xalto.net" <mh@xalto.net>, "nanog@nanog.org" <nanog@nanog.org>
In-Reply-To: <539A1EF7.7090704@free.fr>
Reply-To: Philip Lavine <source_route@yahoo.com>
Errors-To: nanog-bounces@nanog.org
I guess my question is (sorry MPLS speak ahead)=A0with 4 PE's on the edge=
=A0running MP-BGP and ISIS=A0& 4 P's in the core=A0running ISIS, is it best=
practice to confederate or use=A0a route reflector and make=A0the PE's cli=
ents.=0A=0ABasically I want to know what an ISP would do, not a test in a L=
AB.=0A =0A=0A=0AOn Thursday, June 12, 2014 2:45 PM, Michael Hallgren <m.hal=
lgren@free.fr> wrote:=0A=A0=0A=0A=0A=0ALe 12/06/2014 18:39, Valdis.Kletniek=
s@vt.edu a =E9crit :=0A=0A> On Thu, 12 Jun 2014 09:25:20 -0700, Philip Lavi=
ne said:=0A>> need some guidance on best practices=0A>=0A> What the vendor =
says is best practices, or what people in the trenches=0Asay?=0A>=0A>> Is i=
t more efficient to use RR or Confederation?=0A>=0A> If option A is 2% more=
"efficient" than option B, but takes 10% longer to=0A> deploy and causes 3=
% more SLA payouts to your customers when the added=0A> complexity causes a=
whoopsie, how much more work could you have gotten=0Adone in=0A> the time =
you spent in an uncomfortable meeting explaining to upper=0Amanagement=0A> =
why the whoopsie happened?=0A>=0A> (Sorry, it's been that sort of week :)=
=0A=0A:-)=0A=0ANow, Philip, I think along the same path as Vladis: it depen=
ds... What does=0Ayour physical or layer 2 network look like? How do you ex=
pect packets to=0Amove around inside, and in and out, of that topology? You=
need policing?=0AHow much and of what, etc, etc...?=0A=0AI'm quite often a=
fan of confed's, if the network is young thus ``easy''=0Amigration, but th=
ere are scenarios... Please provide more detail to this=0Amail=0Athread or =
one-to-one if you prefer.=0A=0ACheers,=0A=0Amh=0A=0A>
