You're right on and most of us don't monitor for shifts between direct peering and transit very well.  we monitor for bw threshold.   

Paul Bradford

Lead Network Engineer AS11776

C: 814-203-0699

E: pbradford@breezeline.com

 

Breezeline.com 

2875 Rt 764 Suite 2, Duncansville, PA 16635






On Tue, Jul 23, 2024 at 11:23 AM Mark Tinka <mark@tinka.africa> wrote:


On 7/23/24 16:15, Bryan Holloway wrote:

> What irks me is that we have direct PNIs with -- without naming names
> -- the "big guys" delivering this content, and yet the majority of
> this traffic is coming over our public IX connections and transit.
>
> Kinda defeats the purpose of a PNI ... anyone else seeing this?

This issue is fairly common because more network is deployed in
centralized locations that most people can access (like a large city
data centre, e.t.c.), as well as for on-net caches. This is what leads
to exchange point and PNI de-preference switching over to what an
eyeball network may consider "transit".

For the CDN and content folk, exchange point peering (whether PNI or via
the exchange point fabric) is often under-spec'd by consequence, not by
choice. So when clusters or the network feeding them suffers, they tend
to be the first ones to be de-preferred for serving eyeballs. It happens
a lot more often than you may realize, actually, and is not unique to
any one CDN or content network.

Mark.