Re: [Idr] AS hopcount draft as an IDR WG document

In message <Pine.LNX.4.61.0510241445050.21733@netcore.fi>
Pekka Savola writes:
>  
> On Mon, 24 Oct 2005, Elmar K. Bins wrote:
> > I like this proposal very much, because it enables conscious
> > operators to somewhat "confine" their more specific advertisements
> > (or, their advertisements at all) to a certain radius.
>  
> note: I only took a quick look at the document.
>  
> Is this approach really useful?  That is,
>  
>   a) an "AS" is a coarse concept.  A single router, or a single AS 
> could span across the globe.  Thus, "AS hop count" in itself doesn't 
> buy you much.
>  
>   b) given a), this attribute would likely only be used with small hop 
> counts, say 1-3, because if the operator doesn't investigate the 
> AS-level topology in detail, (s)he cannot know which count (s)he would 
> want to use to obtain the intended policy.  I'm assuming investigation 
> of AS-based topology beyond a couple of AS hops would be infeasible.
>  
>   c) as reducing the hop count per traversed AS requires support for 
> this attribute, until this is deployed everywhere, the usage is even 
> more challenging. That is, it's not enough to figure out the AS-level 
> topology, you also need to figure out (and monitor) whether they 
> support AS hopcount, and how to take it into the account in setting 
> the AS level hopcount.
>  
> So, I'm a bit of skeptic whether this kind of tool is useful enough 
> for the intended purpose (i.e., would it become used).  As it happens, 
> you can already use NO_EXPORT for hop 1, and in most cases 
> provider-specific communities for hop 2.  I'm not sure if it's clear 
> enough that a more generic solution is necessary.
>  
> btw. is a confederation peer considered 'an external peer' ?
>  
> -- 
> Pekka Savola                 "You each name yourselves king, yet the
> Netcore Oy                    kingdom bleeds."
> Systems. Networks. Security. -- George R.R. Martin: A Clash of Kings

Pekka,

I think the motivation for this would be the problem that the POISSON
WG set out to address but never found a solution for.

The specific example that was used in POISSON was Australian providers
using a large number of more specific routes advertised to multiple
global providers in order to balance the load from the US west coast
across these providers.  Much of that traffic was from other providers
within the US or from Europe where the more specific routes were not
needed.

Limiting the scope of these more specific prefixes reduced global
routing while still allowing providers to do this sort of load
balancing.  Since we know that this is being done, deployment, even
partial deployment would reduce the amount of unnessecary global
routing information.  Therefore, yes it is useful.

Please feel free to correct me if I got this wrong since I'm not part
of any team that put this together or discussed this in Paris.

Curtis

_______________________________________________
Idr mailing list
Idr@ietf.org
https://www1.ietf.org/mailman/listinfo/idr