Mailing List Archive

asp@uunet.uu.net (Andrew Partan) writes:
* > 04/22/94 18596
*
* AlterNet is currently seeing 19583 routes - looks like ANS is seeing
* about 1K fewer routes. Interesting data point. Not quite sure what it
* means.
*
Hmm. to confirm - this is the data I see from the Table-History file
I am generatiing. It does seem ANS is a 1000 low ?


Date Routes Paths

080494 20136 53243
090494 20139 53225
100494 20044 53333
110494 20036 53183
120494 20294 53642
130494 20326 53849
140494 20374 53822
150494 20246 53541
150494 20384 53776
160494 20099 53167
170494 19949 52708
180494 19858 52489
190494 20238 53577
200494 19327 51054
210494 19463 51341
220494 19486 51544

--Tony.
- - - - - - - - - - - - - - - - -

Andrew:

> 04/22/94 18596

AlterNet is currently seeing 19583 routes - looks like ANS is seeing
about 1K fewer routes. Interesting data point. Not quite sure what it
means.

The following things may contribute to the table size difference:

1. ASs may sent aggregates to NSF/ANSnet and withdrawn more specific
routes from it while still sending more specific routes to other
neighbors. So the route reduced in the NSFNET but not other places.

This is interesting to verify.

2. There are routes has not made it to NSFnet yet. e.g. routes
advertised to CIX usually has a delay to make it to ANSNET.

3. Also, the number 18596 was as of 11:00 am this morning as I indicated
in my previous message. The routing table may grow by the time you
gathered your data. Friday is configuration morning, this behavior
is not unusually.

> The following ASs have withdrawn more specific routes from the
> NSFNET/ANSNet routing table during this period of time:
> 701 AlterNet 112

I think that you really mean that the following peers of NSFNET/ANSnet
(& all of the ASs behind them) have withdrawn more specific routes.

Yes.

If you could identify the home-ASs of the withdrawn nets, that would
probably be more helpful & more accurate.

Actually, I still like to mention NSF/ANSNet neighbor ASs who
withdrawn routes because if they do not do CIDR, AS690 would
not be able to see aggregates from the ASs behind those ASs.
So it is important that a transit AS do CIDR and got recognized.
I will try to do home-AS in addition.
--Jessica

- - - - - - - - - - - - - - - - -

Got a list of routes which are not in the NSFnet table but some other places.
Spot that one incident that the aggregate is advertised to both routing
tables, but there are few specific routes within the aggregate has been
advertised to the NSFNET. So in terms of connectivity, both can reach all the
specific routes destination, but the path would be different. However, that
cause the difference in size of the routing table. This is just the one I
happen to see but there could be more such cases. This indicates
reason #1 in my previous message. I also saw a lot of routes in the diff list
looks like testnets but not quite sure. I do not have time to do more poking
today, I will do it on Monday.

--Jessica
- - - - - - - - - - - - - - - - -

>From list-admin Fri Apr 22 18:19:27 1994
>
>Got a list of routes which are not in the NSFnet table but some other places.
>Spot that one incident that the aggregate is advertised to both routing
>tables, but there are few specific routes within the aggregate has been
>advertised to the NSFNET. So in terms of connectivity, both can reach all the
>specific routes destination, but the path would be different. However, that
>cause the difference in size of the routing table. This is just the one I
>happen to see but there could be more such cases. This indicates
>reason #1 in my previous message. I also saw a lot of routes in the diff list
>looks like testnets but not quite sure. I do not have time to do more poking
>today, I will do it on Monday.
>
> --Jessica
>

I would agree that the difference in table size is most likely due to CIX
issues. I used to see table size differences of a few hundred routes, which
I mostly attributed to the few days of lag time in getting new CIX routes
registered in the PRDB.

ANS-CIX (AS 1957) has a new policy for the past week or two which is
to stop automatically registering the CIX-only routes (routes known
to the CIX which are not registered by another AS 690 service provider)
in the PRDB. I understand they are working out a new policy for getting
these CIX-only routes registed in the PRDB. This could account for some
of the table size discrepency.

Also, as service providers register aggregates in the PRDB, AS 1957
component routes are removed, unless for policy purposes the service
provider requests and ANS agrees to leave the AS 1957 component networks
configured. This could account for several hundred CIX-only component
routes that have recently disappeared from the AS 690 routing tables, but
which may still be announced to the CIX.

Also, there always seem to be a handful of routes (maybe about 20 ?) which
are known to the CIX but which are not registered in the PRDB because they
cannot be verified as having been assigned from address space delegated to
the InterNIC or RIPE.

--Steve Widmayer

- - - - - - - - - - - - - - - - -

>I think that you really mean that the following peers of NSFNET/ANSnet
>(& all of the ASs behind them) have withdrawn more specific routes. If
>you could identify the home-ASs of the withdrawn nets, that would
>probably be more helpful & more accurate.

Thought about this a bit more. If we really want to mention the ASs which have
withdrawn specific routes, homeAS may not give accurate information either.
Think about the case of proxy aggregation, the AS which withdrew specific routes is
the one who does the proxy aggr not the homeAS itself.

--jessica

- - - - - - - - - - - - - - - - -