[as112-ops] AS112 peering coordination?

[William F. Maton Sotomayor]

On Wed, 29 Jul 2015, Aleksi Suhonen wrote:

> Currently the general AS112 peering policy is very open, although only nodes 
> that are hosted at IXPs are easily available for direct peering.
> (Right?)

Up to the AS112 operator really.

> This is alright for peers that only have interconnections in a couple of 
> cities. But for ISPs that have interconnections on multiple continents: I 
> feel it could use better coordination.
>
> A "backbone operator" will be receiving the AS112 prefixes with different AS 
> path lengths and other attributes in different places. They will also 
> probably be receiving it from a mix of customers and peers.

This would be an ideal to achieve, but heard to influence as a volunteer 
group or with $$$ to tie to an end of string and switng over the rail.

> Here's one example view:         http://bgp.he.net/AS112
>
> I think ISPs should forward this traffic to the closest node regardless of AS 
> path length and other attributes. Since that is technically not feasible to 
> guarantee right now, maybe we should try to coordinate peering with "the top 
> peerers," so that they can receive similar announcements on every continent?
>
> Is this something we want to do?

Curious that you bring this up.  At a couple of workshops I had this 
presented:

https://indico.dns-oarc.net/event/21/contribution/31/material/slides/0.pdf

It touches on the issue and gives an example of where the peering is going 
all wrong.  In that I also find - from a loosely inferred point of view - 
Hivane's AS112 node is the hottest on the planet, this despite other 
nodes nearby in Europe and resolvers just passing by them.  It indicates 
that within Europe peering isn't all that great for AS112, or that the 
plethora of IXPs there aren't very effective, or the peering is inadequate 
to AS112 nodes, or ...?

Conversely, the situation is far, far worse on the research networks 
except for a few European countries.  Again, the slides tell all.

But yes, AS112 and peering needs a lot of work for improvement.

wfms