Just some thoughts...
> study peering connect, peering policy and peering reality,
> for the ip carrier of your choice,
> carefully in all markets, which are important for you.
Yes, but this changes daily. If you are single-homed, brace
and hope everything keeps "as is" when things work. Good example:
Swisscom was primary transit was Level(3) for ages and suddenly
changed and brought a new stringent peering policy based on ratios.
It probably changed a lot the perception when the traffic moved
from AS3356 to AS1299 almost in one day.
> I'd say here's where the problem begins and it still UPC network
> I think.
There is no "problem" between UPC and Cogent. This is meant to be so,
because UPC does have a wide footprint and Cogent too and they have
conflicting interests. UPC is known to narrow-minded peering policy.
UPC does have operations in a lot of countries (AT, BE, CH, CZ, DE,
HU, IE, NL, PL, RO, SK) and more to come, they have a lot of eyeballs
and can afford that... the winner here is the one who pays the least
for his transit, or the one who has settlement-free peering with Level(3).
> it is the responsibility of the first network handing over in europe on the
> way "up" and it is the responsibility of the second network handing
> over also in europe on the way "down". (zurich, amsterdam or london
> would be suited in this case)
Correct. Any multihomed network can decide how they want to route.
They control the way, the flow are sent out, but not the way they are received.
Using BGP protocol, it is possible for anybody to influence this and have
the last word, unless depeering gets involved.
> From my experience Swisscom is doing great peering with most
> destinations that I use, be those local or remote (eg the US).
> (the only thing one could say is that some goes over Geneva, but that is
> apparently where their DSL handoff happens)
Not sure they do. Especially with the changes that occured lately. Their
"ratio" peering policy, or paid peering policy will definitely cause collateral
damages. They have already treated to depeer some medium-size network but have
not yet done it. At least for the networks I am aware of. When money is at stake
it seems the good-faith wins ;-)
> My point is:
> PEERING PAYS OFF. No matter how "evli" your competitor might look
> like, both save money and gain speed.
> Even ISP's in other parts of the world have started to realize that
> its stupid to send off traffic across continents just because they
> don't like the competitor.
Your point is very questionable. While peering is a good way to address
route diversification, it also tends to be used to shift "cheaply" all
traffic, ending up with oversubscribed links, higher latency, jitters,
and an overall bad quality. Too many networks see peering as the way to
get masses of bandwidth for cheap, while ending up with suboptimal network
quality of service. Add the hazardous BGP alogrithm for route selection
and you got all the ingredients of a "swinish" network.
Most people who tends to say this lack
Gregory