i saw this one yesterday and was ... uh ... intrigued
does anyone have any idea what data they used?
('top 19 carrier...')
i almost posted it to nanog but questioned
the operational relevance of numbers
that are not differentiated methodologically
from those generated by rand();
[and as such i probably shouldn't responding
to this post...]
quotes:
"We are also seeing that 50 percent of the traffic is being
carried by four of the major ISPs," says Roberts.
Those being AOL (NASDAQ:AOL), MSN, Earthlink, and SBC.
i guess "19 NDA's, sorry, can't say more"
comprises a methodology? that's a nice deal he's got;
most of us can't get away w that...
Roberts' work represents the first hard data collected
on Internet traffic since the U.S. National Science
Foundation discontinued monitoring network statistics in 1996.
hrrrm.... not sure i'm okay with that sentence...
"But, based on what I've seen, the traffic should increase
every six months."
Serious question: What about EarthLink's non-EarthLink-owned modems? (They do lease some modems, I have dialed into one before. And what about NetZero-type companies that outsource all of their modems?
If he is measuring traffic off routers, I do not see how he could differentiate between traffic on modem leased by $PROVIDER_A and traffic on that same modem an hour later leased by $PROVIDER_B.
Also, does he, or anyone else, have any explanation why [insert favorite obvious big backbone] is missing from that list? I know most of the ASes I used to see as "big" are not listed.
> "We are also seeing that 50 percent of the traffic is being
> carried by four of the major ISPs," says Roberts.
> Those being AOL (NASDAQ:AOL), MSN, Earthlink, and SBC.
Serious question: What about EarthLink's non-EarthLink-owned modems? (They
do lease some modems, I have dialed into one before. And what about
NetZero-type companies that outsource all of their modems?
MSN uses at least two dialup wholesale outfits. Earthlink uses their own
modem banks in California, IIRC, and probably also in Atlanta since they
bought Mindspring, and outsources to at least two other companies in other
cities.
SBC probably has its own dialups, but its Prodigy division used to use
Splitrock for dialup access and may still use Splitrock.
If he is measuring traffic off routers, I do not see how he could
differentiate between traffic on modem leased by $PROVIDER_A and traffic on
that same modem an hour later leased by $PROVIDER_B.
Post-merger Earthlink uses a combination of outsourced modems and large
aggregation POPs for dense regions, depending (obviously) on the financial
layout for any given area.
This isn't NDA info; any customer can generally find out whether they're
dialing into one of the super-POPs or an outsource by looking at their
own reverse, and/or asking support where the number leads to.