As Bill Woodcock asserts, there is no question that peering at exchanges is
driven by economics more than other advantages such as path minimization
or performance maximization, whether you are peering amongst Tier1's or are
a Tier2/other minimizing your transit $$$ by maximizing your peering
traffic %.
But the drop in transit pricing, and things like burstible billing, have
definitely changed the "margin" in this calculation. So now we see folks
going to exchanges/neutral colocation for much more than just peering
arguments, but a combo of:
-public peering
-private peering
-access to lots of low-cost transit providers
-access to lots of low-cost circuit providers--including
ethernet "local loops"
-ease of switching bandwidth suppliers
-just need a good facility to put a POP
-sell services
As for IXen in Europe versus US, the history (having suffered thru this at
MAE-East) is that FDDI-based US exchanges reached congestion first, and the
largest traffic pushers were forced to go the private peering route. So
legacy exchanges then entered a rather spotty period of faltering critical
mass, competing exchanges, diverging technology choices (ATM versus Ether)
and a mixed private/public interconnection model. Today, congestion is not
a problem at US IXen, but having critical public peering mass in all the
major geographic peering areas is.
Whereas European public exchanges never experienced congestion, since by
time
traffic growth hit, there were egress-buffered GigE switches available.
Thus
they continued merrily along toward critical mass, and traffic at exchanges
like LINX (surpassed 13 Gbps this week!), AMS-IX, and even Japanese
exchanges
have surpassed legacy US exchanges. However, new entrants to RIPE and APNIC
IX business may certainly pull at the peering concensus in these areas.
Cheers,
-Lane