[Fwd: Unstable BGP Peerings?]

Christopher Morrow morrowc.lists at gmail.com
Wed Jan 16 03:12:24 UTC 2008


On Jan 13, 2008 6:56 PM, Paul Ferguson <fergdawg at netzero.net> wrote:
>
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Interesting, given that TTNet sits atop this ranking:
>
> https://nssg.trendmicro.com/nrs/reports/rank.php?page=1
>
> I wonder if this is somehow related? ;-)
>

probably not... but only based on as much a guess on my part as paul's
I suspect, a little more below.

> - -- Sue Joiner <smj at merit.edu> wrote:
>
> Forwarding for Mohit Lad and Jonathan Park.
>
>    -sue
>
> Sue Joiner
> Merit Network
>
> - -------- Original Message --------
> Subject:        Unstable BGP Peerings?
> Date:   Sun, 13 Jan 2008 17:49:44 +0000
> From:   ParkJonathan <j13park at hotmail.com>
> To:     <nanog at merit.edu>
> CC:     <smj at merit.edu>, <mohit at cs.ucla.edu>
>
>
>
> During our recent study on BGP routing instability, we found cases where
> lots of routes changed from one subpath to another subpath, repeating this
> kind of behavior over a few months . We do not know the cause of this
> repeated instability, but suspect the BGP peering between routers in two
> AS was unstable or had some problems and this caused routing changes seen
> by many observation points.

It seems, to me, that from the data you have on the website perhaps
this is just oscillation in best-path decision or internal traffic
engineering decisions exposed to the outside world? Perhaps (taking
the first picture example) 9121 decides partway through a day or month
that they want to use cogent more (ratio levelling or cost reasons) so
they draw traffic via 174, then after some metric is met (cost or
ratio) they spread the load across their other transit links? This
could easily account for the changes you are showing, right?

In point of fact the next few examples also seem to reflect the same
behaviour... I suppose this could even be automated with one of those
fancy-dan InterNap route-optimizer boxes, right? I'd be curious to
know if this makes sense to: 1) other folks on-list, 2) the
researchers.

-Chris
>
> Specifically, the peerings in question are
>
> 174 (Cogent) and 9121 (TTNet)
> 3257 (Tiscali) and 9121 (TTNet)
> 9304 (Hutchison) and 15412 (Flag Telecom)
> 1273 (Cable&Wireless)  4651 (Thai Gateway)
> 6762 (Seabone) and 7473 (Singapore Telecom)
>
> For details of events and timings, please find a short summary on the
> link below.
> http://irl.cs.ucla.edu/pca/active-links.html
>
> We would really appreciate if somebody from the ISPs involved in this
> activity (or anybody who might know what happened) would contact us and
> throw some light on the reasons for this behavior.
>
> Thanks
>
> Mohit Lad, Jonathan Park
> UCLA
>
> -----BEGIN PGP SIGNATURE-----
> Version: PGP Desktop 9.6.3 (Build 3017)
>
> wj8DBQFHiqUVq1pz9mNUZTMRAkVEAJ0Y4u5AYr/CiG65e3e+Y88HCQJGjQCg723O
> P17FTAUFOw3Ms1KQW6v2+44=
> =013x
> -----END PGP SIGNATURE-----
>
>
>
> --
> "Fergie", a.k.a. Paul Ferguson
>  Engineering Architecture for the Internet
>  fergdawg(at)netzero.net
>  ferg's tech blog: http://fergdawg.blogspot.com/
>
>
>



More information about the NANOG mailing list