Re: [apops] Re: The Mandatory One Reply To This Weeks: The Cidr Report

["Phillip Grasso-Nguyen" <phillipg@magna.com.au>]

Not if backbone providers allowed users to control their own routing
policies via communities,
This wouldn't be hard, but what it does mean is that it gives much more
control to end providers which direction traffic flows.

e.g. if matching a community (small example). This would stamp the route
apply these settings to it's directly connected peers.

5511:10000, set no-export to all peers
5511:10001, set no-export to AS701
5511:10002, set no-export to AS6461
5511:10003, set no-export to all European peers
5511:10004, set no-export to all north american peers
5511:10005, set no-expor tto all Asian peers
5511:11000, set prepend once (5511) to all peers
5511:11001, set prepend once (5511) to AS701
5511:11002, set prepend once (5511) to AS6461
5511:11003, set prepend once (5511) to European peers
5511:11004, set prepend once (5511) to north American peers
5511:11005, set prepend once (5511) to all Asian peers

blah blah, i think you would get the hint, if large backbone providers did
this it will give people enough flexibility to set control traffic
direction/flow whilst still maintaining a CIDR only routes. What about
control, it's easy, sh ip bgp community 5511:10000 to see who's doing what.
if you have an community-list 5 matching 5511:11005 then show ip bgp
community-list 5

Regards
    Phillip
-----Original Message-----
From: "Joe Abley" <jabley@automagic.org>
To: "Christian Nielsen" <cnielsen@exodus.net>
Cc: "Hank Nussbacher" <hank@att.net.il>; "Aleksi Suhonen"
<nanog-poster@axu.tm>; <apops@apnic.net>
Date: Thursday, 22 March 2001 10:07
Subject: Re: [apops] Re: The Mandatory One Reply To This Weeks: The Cidr
Report


>On Wed, Mar 21, 2001 at 02:43:10PM -0800, Christian Nielsen wrote:
>> On Wed, 21 Mar 2001, Joe Abley wrote:
>> > The solution of advertising a large number of prefixes in different
>> > ways to different adjacent networks is one way of accomplishing the
>> > policy dissemination; however, one problem is the large amount of
>> > prefix bloat associated with it that the whole network gets to see.
>>
>> right. that is correct. i guess i can better understand this if they, the
>> end user/edge network, would run BGP.
>>
>> *>i148.182.0.0      209.1.40.63                  1000      0 1 16779 1221
i
>> *>i148.182.16.0/24  209.1.40.63                  1000      0 1 16779 1221
?
>> *>i148.182.17.0/24  209.1.40.63                  1000      0 1 16779 1221
?
>> *>i148.182.18.0/24  209.1.40.63                  1000      0 1 16779 1221
?
>>
>> i have no problems seeing
>>
>> * i148.182.12.0/22  209.1.220.156                1000      0 5727 1221
4740 4740 4740 4740 i
>
>I think the issue here is you are only seeing part of what 122 may be
>trying to do by your single view. If you can imagine other views of the
>network where those /24s have different attributes, you can see how
>there might be some method behind the madness; the deaggregation in
>that case is a method for 1221 to influence routing in remote ASes in
>order to do it's own inter-AS traffic engineering.
>
>Note that I have no real knowledge of what 1221 is doing, or why. However,
>Geoff is in Minneapolis presenting on exactly these topics and it seems
>entirely possible that the two observed phenomena are related :)
>
>
>Joe
>*             APOPS: Asia Pacific Operations Forum              *
>* To unsubscribe: send "unsubscribe" to apops-request@apnic.net *
>

*             APOPS: Asia Pacific Operations Forum              *
* To unsubscribe: send "unsubscribe" to apops-request@apnic.net *