Re: a different proposal for ipv6 in bgp

Dimitry Haskin <dhaskin@baynetworks.com> Tue, 31 December 1996 03:07 UTC

Received: from cnri by ietf.org id aa09599; 30 Dec 96 22:07 EST
Received: from merit.edu by CNRI.Reston.VA.US id aa24201; 30 Dec 96 22:07 EST
Received: (from daemon@localhost) by merit.edu (8.8.4/merit-2.0) id VAA04963 for idr-outgoing; Mon, 30 Dec 1996 21:37:19 -0500 (EST)
Received: from interlock.ans.net (interlock.ans.net [147.225.5.5]) by merit.edu (8.8.4/merit-2.0) with SMTP id VAA04958 for <bgp@merit.edu>; Mon, 30 Dec 1996 21:37:16 -0500 (EST)
Received: by interlock.ans.net id AA24594 (InterLock SMTP Gateway 3.0 for bgp@ans.net); Mon, 30 Dec 1996 21:37:15 -0500
Received: by interlock.ans.net (Internal Mail Agent-3); Mon, 30 Dec 1996 21:37:15 -0500
Posted-Date: Mon, 30 Dec 1996 18:36:42 -0800 (PST)
Received: by interlock.ans.net (Internal Mail Agent-2); Mon, 30 Dec 1996 21:37:15 -0500
Received: by interlock.ans.net (Internal Mail Agent-1); Mon, 30 Dec 1996 21:37:15 -0500
Message-Id: <2.2c.32.19961231023343.00740e3c@pobox.corpeast.baynetworks.com>
X-Sender: dhaskin@pobox.corpeast.baynetworks.com
X-Mailer: Windows Eudora Pro Version 2.2c (32)
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Date: Mon, 30 Dec 1996 21:33:43 -0500
To: Tony Li <tli@jnx.com>, dhaskin@baynetworks.com
From: Dimitry Haskin <dhaskin@baynetworks.com>
Subject: Re: a different proposal for ipv6 in bgp
Cc: bgp@ans.net
Sender: owner-idr@merit.edu
Precedence: bulk

At 05:59 PM 12/30/96 -0800, Tony Li wrote:
>
>   I can't say for all, but for some ISPs the size of BGP policy database
>   way exceeds the size of all other bgp data structures including RIB and
>   hence is a real issue. 
>
>That's a very interesting claim and I'd like to understand it in more
>detail.  Could you precisely define what the "policy database" is?  Are we
>just talking about the administrative filters?

Yes, the "policy database" for me is what is the "administrative filters"
for you -- the same thing.

> Or are we talking about database of paths?

And this too.  

> If the latter, are AS path strings separated from the RIB?

In the implementation that I'm familiar with, they are (probably the
true answer depends on how we define the RIB boundaries).
Also there are the regular expression storage which clearly
sits within the policy database.

>
>In one particular implementation that I'm familiar with, the dominant
>memory consumer is the RIB, and that's by several orders of magnitude.  In
>that implementation, the AS path strings are separated from storage in the
>RIB.
>

It is not my experience (probably mach more limited than yours ;).  I guess it
depends how extensively a particular ISP uses the "administrative filters".

>Tony

Dimitry