[30235] in North American Network Operators' Group
route-reflector hierachies
daemon@ATHENA.MIT.EDU (Lavelle, Neil)
Thu Jul 20 10:35:20 2000
Message-ID: <CB1E59E84CE5D3118E5C00508B6D755551378E@s1000.orchestream.com>
From: "Lavelle, Neil" <nlavelle@orchestream.com>
To: "'nanog@merit.edu'" <nanog@merit.edu>
Date: Thu, 20 Jul 2000 15:31:04 +0100
MIME-Version: 1.0
Content-Type: text/plain;
charset="iso-8859-1"
Errors-To: owner-nanog-outgoing@merit.edu
Hi All,
I have a debate raging with one of my co-workers as to how one should deploy
route-reflectors that has reached the point where I need some help from
people that know lots about it ;-)
If I have a number of clusters each with its own route-reflector(s) how
should I peer between clusters ?
From reading RFC1966 ( and common sense ) it seems that the clusters should
be fully meshed, and this is my opinion, however another opionion is that
one should build a tree of route-reflectors with some route-reflectors not
serving clusters but solely existing to reflect to other route-reflectors.
Now I think the second idea would work and might be useful in some very
specific cases, but would be madness given that each of the route-reflectors
that form nodes of the tree represent single points of failure and that a
failure of the root node would have catastophic results ....
Any comments ?
Thanks in advance
Neil
--
Neil Lavelle
Orchestream Avon House
Tel: +44 (0)20 7348 1500 Kensington Village
Fax: +44 (0)20 7348 1501 Avonmore Road
http://www.orchestream.com/ London W14 8TS, UK
