All, the following was sent to me by John Stichnoth on the Gilder thread:
Message 8327854
>>Just to clarify.... My understanding is that the telcos can run T1 from
businesses to a central location, then bundle several into one T1 router
"at the edge" of the Internet. <<
John,
There are a growing number of delivery schemes which deliver T1
speeds. There are the original T1 lines which were first implemented in
1962 over copper pairs. There are those which use fiber and wireless
facilities. These operate at a line rate of 1.544 Mb/s using what is called
the DS1 format for the most part (which is easily broken down into 24
individual 64 kb/s voice channels, plus an 8 kb/s overhead channel).
Then there are digital subscriber lines (DSLs) that can deliver ~1.5 Mb/s
in the downstream direction and considerably less (64 to 680 kb/s) in the
upstream. There are also HDSLs and SDSLs that deliver 1.5 in both
directions, symmetrically.
My point here is that all of these are priced differently, and they each
require different levels of provisioning and administration by the carrier or
SP. More to the point, however, is that each of these line types can fit the
scenario you've inquired about. Therefore, I can't give you a general
answer that will fit all cases.
What you say is true. Multiple T1s will be aggregated by a carrier or ISP
at the edge of the network into a single router. On the other side of the
router there will invariably be a fatter pipe [or combination of pipes (each
going to other router sites)] of some statistically derived speed.For
discussion, consider a T3 line in this case, operating at 45 Mb/s that will
be used by all of the individual T1s. There could be many, into the
hundreds, of T1s on the subscriber side, and only 1 T3 on the core side
of the router.
From this you can see that if you had 100 T1s (totaling >150.0 Mb/s),
they could easily overwhelm a single T3 (45 Mb/s) if they were all active
at the same time. Statistically, this doesn't happen, according to the
behavior of most consumer and small business traffic patterns.
Nonetheless, if a majority of the T1s were to go live at the same time, it
would slow down each subscriber's performance, considerably. The
congestion at this point would be in the router, or in the edge domain, or
in the upstream trunks leading to the core a NAP. The sizing of each of
these network elements would play in this calculus.
The oversubscription ratio, that is, the ratio of users (the T1s) to
backbone access capacity (in this case the the T3) is an important factor
in determining the probability of congestion-affecting slowdowns in the
edge.
>>If a small business is getting T1 at much less than a couple thousand
dollars, they are likely sharing bandwidth with others. (Which might be fine,
if they are all small businesses who don't use it all up).... Is that correct?<<
Yes. Internet access, by definition, assumes that they will be sharing
bandwidth with others. And yes, this does result in a reduction in charges
to the individual.
HTH, Frank_C. |
