There is a non-trivial amount of WAN traffic flowing between ISPs' private
interconnects. They do not touch NAPs at all. I wonder if this factor
is taken into consideration when the 1/3 figure was generated.I'd tend to doubt it.
So you do not think that the estimate include the traffic exchanged at
private interconnections either. If this is indeed the case (which is
very likely), the figure is not that meaningful.
Do you know any way to get flow stats from every backbone collected in one
place and collated?
I'd think most of the backbones have stats data of its own private
exchange(s) but I do not think it's publicly available.
--Jessica