Curtis, and Thread,
I'm in agreement that the larger providers, certainly the newer ones on the scene like QWST, and the many who are now forced to keep pace with them, are making strides to dramatically increase their overall "capacity" by whatever order of magnitude they can, and in some cases through whatever means they can. Anyone who reads press releases and who can add and multiply using DWDM factors, will agree with this, if not in real time actuality, then at least in the trend.
I simply don't now how such a claim as "throughput" doubling every 100 days [which extends to a ten-fold or greater annual nut to swallow] can be substantiated in such actual, precise metrics, or even approximated to within a reasonable margin of error.
That's essentially why I've posted this question, upstream.
I think that it is unlikely that such a claim is at or anywhere near reality, despite the claims being made by our favorite magazine columns and pundits du jour, or the providers [both SPs and vendors, alike] who would like to be seen in this light.
Right now I believe that it is an overstatement. In the future, we may go through periods when it is actually an understatement. Such persons as a Robert Metcalfe can't be wrong all of the time. Sometimes their timing is just off.
The "capacity" is increasing at this rate, possibly, in some cases. But I have yet to see where anyone has implemented multiple DWDM-32's or 40s or 80s in one fell swoop, which is what it would take to make even the capacity metric hold true.
I have reservations about this, too, because capacity is defined in many different ways, and many of the lambdas that are now being put into "potential" are being deployed in reserve, i.e., often unlit, and even some of the lit ones are in back-up roles to one another, and not necessarily for supporting primary throughput functions. Although, the majority of those which are "actaully being lit" are going to live traffic, which only makes sense.
I think that the pundits and others who must say sensational things in order to make a living have taken license here and used this new found capacity "potential" to define in large measure what they are now calling "throughput."
I have surmised, whether correctly or not, that as soon as a switch is thrown on a WDM increasing capacity by whatever measure, actual congestion levels and throughput loads see a sustained period of "decline," in fact.
This is true because at that moment in time, and for some sustained period beyond, there is less congestion on those routes, and there are fewer cases of time outs and restarts. It's not that simple, I know, but that effect cannot be discounted, however short lived in terms of weeks, months or quarters it may be. Notice, I've stopped short of suggesting years.
My main point here is that IMO throughput increases do not immediately and directly track with capacity increases -- at least not within the immediate time frame. There are no "starter-gun" dynamics at play here, and there are too many other variables at play to be dependent on, not the least of which are the politics and the business environments at the NAP- and Tier One- levels.
If actual traffic were increasing at the same rate as capacity, then there would never be any noticeable gains in performance improvements [which in reality there is], for that would mean that the newly-made-available-head-room would be consumed as fast as it were being placed.
There is more often than not a lag that takes place, though, sometimes a long lag as witnessed by improved and sustained response times being afforded by many providers today, despite the increases in traffic of whatever magnitude.
The momentary improvements that were made available at the ATHM maiden location when it first went live, and the ensuing situations of the current day, serve to personify these dynamics, IMO, despite there being some distortions regarding exactly what the actual causes there have been, there.
The area that is least clear to me at this time is that of the VPN, and extranets, and how these affect the overall. NAPs can do weather reporting of public 'net activities, but the often do not reveal those partitioned v-c's and redirected flows that the ISPs use for their corporate accounts. Some accounts actually have this form of non-disclosure written into their contracts.
This may be an area that is increasing faster than the stated rising tide every 100 days, or slower, I don't know. And I don't know if anyone else does, either.
I will grant you that in time, perhaps in "short time," all of the increases in capacity will become consumed, however, and the need for escalating capacity will continue. And so it goes.
|
