Re: BGP4 stuff: Local Preference Computation
Curtis Villamizar <curtis@ans.net> Wed, 04 September 1996 23:34 UTC
Received: from ietf.org by ietf.org id aa12178; 4 Sep 96 19:34 EDT
Received: from cnri by ietf.org id aa12172; 4 Sep 96 19:34 EDT
Received: from merit.edu by CNRI.Reston.VA.US id aa18237; 4 Sep 96 19:34 EDT
Received: (from daemon@localhost) by merit.edu (8.7.5/merit-2.0) id SAA07068
for idr-outgoing; Wed, 4 Sep 1996 18:55:23 -0400 (EDT)
Received: from interlock.ans.net (interlock.ans.net [147.225.5.5]) by
merit.edu (8.7.5/merit-2.0) with SMTP id SAA07063 for <bgp@merit.edu>;
Wed, 4 Sep 1996 18:55:20 -0400 (EDT)
Received: by interlock.ans.net id AA20691
(InterLock SMTP Gateway 3.0 for bgp@ans.net);
Wed, 4 Sep 1996 18:55:18 -0400
Received: by interlock.ans.net (Internal Mail Agent-1);
Wed, 4 Sep 1996 18:55:18 -0400
Message-Id: <199609042254.SAA07748@brookfield.ans.net>
To: Cristina Radelescu-Banu 617/890-1001 <cristina@midnight.com>
Cc: bgp@ans.net, rwoundy@vnet.ibm.com, jgs@ieng.com
Reply-To: curtis@ans.net
Subject: Re: BGP4 stuff: Local Preference Computation
In-Reply-To: Your message of "Wed, 04 Sep 1996 15:34:29 EDT."
<199609041934.PAA00467@rasarit.midnight.com>
Date: Wed, 04 Sep 1996 18:54:46 -0400
Sender: ietf-archive-request@ietf.org
From: Curtis Villamizar <curtis@ans.net>
X-Orig-Sender: owner-idr@merit.edu
Precedence: bulk
In message <199609041934.PAA00467@rasarit.midnight.com>om>, Cristina Radulescu-Ban u writes: > > I have some problems in implementing BGP4, and if you could > clarify some of them , this would be great. > > I understood that LOCAL_PREF is used as degree of preference only for > the BGP4 speakers located in the SAME Autonomous System. The Degree of That's MED. LOCAL_PREF has always been compared regardless of AS. > Preference is something totally different, it is used both for the BGP > Peers in the same and in the neighbor AS (it can use in its > computation the LOCAL_PREF, that's true), and one of the devices I > tested computes the Degree of Prefence with formula: > > TotalASPolicyWeight+ PeerWeight + DefaultWeight > > where DefaultWeight = the weight added to each route when computing > the degree of preference for the route (LOCAL_PREF)-if you have > any-because it's an discretionary attribute-, and > if you consider it because it was received from a peer in the > same AS) > You can assign weight to each AS in the AS_PATH => this way you have th > e > TotalASPolicyWeight; > > PeerWeight = the weight that is added to all routes received from the specifi > ed peer You can assign LOCAL_PREF any way you want. Rich just made a suggestion since preferring shorter AS paths if you have no other way to decide which route is better is a popular idea in some circles. > For me this is pretty clear, to take in account > the AS and the Peer when you compute the Deegre of Preference. > > After the router has done the computation for this Degree of > Preference, the Local Pref assigned to the route take the value of > the Degree of preference already computed. If an internal peer > received that route, it may or may not > take as Local Pref the LOCAL_PREF already received (it can compute > another Degree of Pref based on this Local Pref); If an external peer > received it, it ignores it and does the Degree of Pref computation for > the route again, without taking in account any Local Pref. > > Yes, sometimes the Local Pref can be computed (if the internal > peer choose this)but the one which is computed is always the DEGREE OF > PREFERENCE, which is different form Local Pref. > > This is totally different from what I saw in the messages > exchanged, about the Local Pref Computation, which I think should be > the Degree of Preference Computation; what is the right thing to be? > > Thanks in advance, > > Cristina IMO Yakov has obscured the use of LOCAL_PREF by referring to degree of local preference when "value of the LOCAL_PREF attribute" belongs in the draft. This might be left over from early use of BGP4 where some people didn't set local-pref at all. Curtis
- BGP4 stuff: Local Preference Computation Cristina Radulescu-Banu
- Re: BGP4 stuff: Local Preference Computation Curtis Villamizar