You might recall the Cable Hotels I referred to earlier. These would resemble (or be the same) Carrier Hotels that serve as colocation facilities for CLECs and other common carriers. Well, these would actually serve as cablecolos too, or could materialize in various other ways.
The backbone providers, being the mountains in this case, would, in all likelihood, find opportunities in delivering capacity directly to those colo's, as they do today. Such colos contain the presence of upper-tier providers. The higher the density, the more Tier 1 thru 3's you'll find. Sometimes, the T1s, T3s, or whatever, to the core amount to nothing more than a local cross connect on site within the colo.
The "meet" could take place either in large parcels of real estate in, or adjacent to, existing large head end sites, or remotely situated colos, such as I've been referring to, connected by "back-hualed" fiber connection to the head ends.
These approaches to connectivity are consistent with current ISP practices today in the normal residential dial up and commercial dedicated line configurations.
Some of these colos are local meet points, some much larger, mimicking NAPs, or private NAPs, and others are merely convenient hubs that happen to fall in the nexus of common carrier routes in busy commercial districts. [Do you suppose that Gore will propose that libraries and public schools become the next colos for the cable industry to home in on?]
In these cases, fiber optic "meet-me" rooms could be established in the cable operators' head ends, and the traffic simply hauled back to the colo via fiber. In this fashion, a "virtual" presence would be created at the head end, because of the sense of transparency afforded by the high speed lines. And, as the costs for bandwidth continue to decline, the sense of transparency increases.
Here we are talking about ILEC or CLEC carrier fiber doing the backhauling, actually SONET-derived T1s, T3s, etc., for the most part (although I think that in this new model we might actually start to see some IP over lambda soon) that would be used to carry the ISP's IP traffic back to the colocation site. At the colo site, it would very likely meld with normal dialup traffic, and forwarded to the core.
The one inescapable thing that any interloping ISP would need to contend with, however, unless they were to ride over someone else's flows in a subordinate role, is the process of inserting and extracting their traffic at the actual head end at the RF stages, and receiving/forwarding said traffic to their own routers. Again, it is conceivable that smaller ISPs may cut deals to ride over the flows of larger players, which would not be unusual or inconsistent with the Internet approach to doing business.
In turn, the traffic would then need to be sent to their own upstream providers, whether directly from the cable operator's head end, or from the colo. Of course, this all adds to the burdens of the cable operator, for which they would need to be compensated. I have yet to see anywhere someone proposing how the logistics of this would play out.
Agreed, it would seem that the number of T1s (1.5 Mb/s) for the smaller ISPs and the number of T3s (45 Mb/s) and OC-3s (optical carrier level 3, or, 155 Mb/s) required to do this would be very high and unwieldy.
In some instances, this would actually allow some of the Interloper Service Providers an opportunity, albeit over time, to do some house cleaning, i.e. very gradually re-grooming their networks, by consolidating disparate T1s into larger T3s, reallocating load handling, etc. For the most part, however, it would be a royal pain in the ass. But, as someone already asked, 'what price speed?'
The above "gradually" qualifier is very important to grasp here, because such a transition would not occur overnight, taking years to accomplish, before some equilibrium in the mix was reached.
Regards, Frank Coluccio |
