Re: ISP Policies

[Stefan Mink <mink () schlund net>]

On Thu, Sep 09, 2004 at 05:10:21AM +0000, Christopher L. Morrow wrote:
> I'm not a router guy (routing atleast), but perhaps there are performance
> problems inside an ASN along a path which you connect to other places? So
> you might lengthen paths through/to that ASN to force traffic across
> another ASN's direct connection which is less problematic?
>
> Or, you just don't want to send traffic through Bill Manning's ASN because
> you dislike his hawiian T-Shirt Policy? There are probably a few hundred
> reasosn why you'd avoid an ASN... In general though I'd think that like
> Michel said: "It's a pain and its doing something that bgp should do for
> you without lots of messing about"

choosing a route whose AS path doesn't contain an AS is no garantee
that packets won't flow through that AS nevertheless. Forwarding is
being done hop-by-hop. Examples for that szenario are
 * more specifics that aren't distributed globally
 * broken/complicated routing setups where a router chooses a route it got
   via IBGP and where on the path to the corresponding nexthop another
   router decides to rather use his own route (e.g. a router chooses his
   own route due to step "external vs.  internal" in the BGP decision process, 
   but the original decision was based on lowest router id) This can happen
   in not-full-mesh environments, e.g. with route reflector setups...

I use the AS-path rather to grab chunks of prefixes (e.g. all routes
via ASxyz) to do loadbalancing over transit providers whose announcements
look so similar that the decision is only based on lowest router id of the
IBGP speaker. I know this can be "somewhat solved" via multi-path bgp but
there are other issues with that (fib size, multiple exit points behind
one route-reflector etc)...

   tschuess
             Stefan
-- 
Stefan Mink, Schlund+Partner AG (AS 8560)
Primary key fingerprint: 389E 5DC9 751F A6EB B974  DC3F 7A1B CF62 F0D4 D2BA

Attachment: _bin
Description: