AS number consolidation

Fri May 30 14:03:54 UTC 2003

SBC just did one of the largest AS merges ever accomplished. Maybe one of
their engineers will give a brief report? One item to always look out for,
if you have BGP-speaking customers is the particular implementation of
"local-as" on your favorite brand of router. There are some variations...

- Some include the "imposter AS" in the AS path
- Some have a keyword for stripping out the "imposter AS"

The vendor documentation on this feature has been historically weak,
necessitating some lab work.

This can be an issue, as you could be accidentally prepending route
announcements learned from your downstreams by having both the new and old
AS in the path.

Thanks,
Daniel Golding

On Fri, 30 May 2003, David Luyer wrote:

>
>
> > Does anyone know of case studies of companies collapsing
> > multiple ASes
> > into one on their network? I have the Allegiance Telecom
> > presentation from
> > NANOG 27 but I would like to hear how other people have done
> > it as well.
>
> We have to date collapsed 6 AS numbers into 1.
>
> Approach was relatively simple - but it's been at
> least a couple of years since the last merge so I'm
> a little light on exact memory/detail:
>
>   1.  Duplicate RADB (or other) entries across to new
>       AS.
>
>   2.  Merge interior routing protocols across the ASen
>       which are about to be merged.
>
>   3.  Gradually grow the largest AS adding a router at
>       a time (notify the BGP peers on the router, set a
>       time, make the change).
>
> To me this is one of the things where you can go over the
> top in planning (I know an ISP who have been planning an
> AS merge for years) or you can 'just do it'.
>
> The second hardest thing, if you bother to do it, is
> "cleaning" the old AS (in terms of RADB, other registries,
> route filters in peers, etc) before returning it to the
> registry.
>
> The hardest thing is convincing the registry to take them
> back... (you'd think that'd be easy, but it took by far
> the longest of the whole job).
>
> David.
>
>