Choosing Peering vs Transit

List - I’m curious as to how one chooses to Peer versus buy Transit in the Internet. What are the determining factors, if I were to start an ISP, on which I would do to get my network recognized by the Internet and how I would get my customers out into the backbone.


Hmm..If you were (or are) starting an ISP, more than likely you
will be buying transit. You won't meet many of the larger ISPs peering
agreements. For example UUNET (Or Worldcom, whatever they call themselves
today, maybe it will be UUcom soon!) requires an OC-12 backone, presence
in 15 states, ability to peer at OC-12 level at atleast 4 points, and you
must already be at 3 of the following MAE-EAST, MAE-WEST, AADS,
MAE-CENTRAL and peered with them. Most other ISPs are similiar in these
requirements, however, some are more open than UUNET/Worldcom.

  There are some advantages to buying transit, however. Obviously
there is a larger cost incurred with transit but you get SLA's, more tech
support, ability to tune routes with metrics/etc. Additionally you might
only need a DS3 to some providers while you need more to others (OC3+).
So in reality buying transit could give you more control to use your
bandwidth to larger ISPs more effectively.

  In the beginning you should look into a mixture of public peering
at various public peering points and perhaps some strategic purchase of
transit. Times are hard in the Internet right now. Honestly a big factor
(Atleast in my mind) in going to one backbone or another is cost and
performance. Additionally you need to think about what your backbone will
be providing. If you plan on mainly doing dialup, cable modems, lower end
providing sometimes just getting packets from A too B is more important
than adding an additional 10-20ms on each of these packets. If you are
doing more of a performance based ISP, higher speed customers, content
hosting, etc, the previous statement may (and more than likely will not)
hold up. Customers will not be happy with a trans USA path being
100-120ms instead of 60-70ms.

  In reality you need to think about your game plan. You just can't
go out and say, "Hey I'm building an ISP", and get a blanket answer.
There are many factors that should determine the approach you take. Of
course that's just my opinion. I could be wrong.