No subject

David Diaz techlist at smoton.net
Thu Dec 19 02:42:40 UTC 2002


I have to disagree with you.  I dont think most of their customers 
are adult content or the likes looking for cheap transit.  I think 
there model was to attract business users that wouldnt use the full 
pipe.  They can burst but arent expected to saturate the pipe 24x7. 
It's not a bad model, but it would need time to develop.  They did 
buy a lot of peering, but as we know, you cant really do that. 
Peering is not a permanent assignment.

What may have happened is that they did get a few content providers 
that then became a large amount of their backbone when looking at 
traffic levels, not number of customers.  I do know they really tried 
to balance the outbound with significant inbound.  They even went so 
far as to have all gigE sales have to get approved by engineering.

As for AOL, they have a lot of content or two way traffic, depending 
on who is on your backbone.  They have a lot of hosting and content, 
and let's not discount peer to peer traffic.  It was interesting to 
look at peering connections btw eyeball to eyeball networks. 
Shocking how much traffic there was.

The issue may have a bit to do with the fact they had a lot of 
different backbones they had to integrate and manage, many AS#s and 
different peering pipes and relationships.  It was a complicated 
process and their peering people did a good job.  But they let them 
all go.  So ones the chef(s) were out the door, who maintains it? 
It's not exactly a status quo situation.  It was treated in house as, 
well we have the peering so why do we need all these pain in the butt 
expensive guys?  This is what I heard and it seems to be accurate.

In short order they started having traffic issues as well as peering 
relationship issues.  To be fair, it's hard for engineers on the 
other side of the peering session to work with a blackhole.  The 
stability of their network is depending on the quality of the 
engineers of their peers.  Without those key people to work with at 
Cogent it seems some peers are deciding it's not worth the risk.

It is ashame because there are a lot of companies using cogent to 
build themselves backup in a bad market.  The price was great and it 
was helping to revive a bad space.  This just puts them into panic 
mode, and Im not sure they truly understand what the issues are. 
What they will do is call both cogent and aol and complain and feel 
bitter.  Not exactly our finest hour.  Another outage without prior 
warning.




At 21:17 -0500 12/18/02, Ringdahl, Dwight (WebUseNet) wrote:
>>>  Thing is if your connection is completely full one way, it'll effect
>>>  traffic the other way too.
>
>My thoughts are Cogents primary customers are sites that are looking for
>very cheap bandwidth, which most likely is adult content. Therefore they
>would look more like a content provider than a transit provider.
>
>My question, being a content network, is how would AOL expect them to have a
>balanced pipe? AOL is all eyeballs, and really doesn't have much content
>which is useful outside of their user base. Especially if you already peer
>with Time Warner in other sites.
>
>When ISP's peer I would have thought it is to prevent having to pay transit
>companies like Level3 for the bandwidth. This leads me to believe there
>might be something more to this, like maybe the spam spewed from the adult
>sites. Just a guess, anyone have any hard data?
>
>Dwight





More information about the NANOG mailing list