Hi Bill,
You got me! Touch‚!
>>But we're talking about data, asymetrical traffic flows, "split" lines, spectrum management and heterogeneous layer 2/3 equipment. What "larger" carrier do you know that doesn't have the same traffic engineering problems that smaller ones have in this type of environment?<<
I once wrote a rather lengthy SONET article, fashioned around a form of tutorial, for one of the popular trade mags. The editor, at the time, asked me to go out on a limb and make some borderline statements on the side of the bleeding edge, in order to stimulate discussion and feedback. I guess he was right. <smile>
You'll note my use of the (<??>) in my post, which should have been a dead give-away that although I know there is a performance bogie that the carriers will have to live with, and not fall short of, I don't know exactly what it is. But such things don't go unattended to in the top tier, or world class carriers. Execution, however, may be another story, which is what I think one of the points of your post was about, and I agree.
Where they size their true ADSL lines (that employ splitters), some of the engineering has already come into play to support voice objectives, but to what extent this actually assists in the sizing of the data side, is beyond me, although my guess would be very little, if any at all.
And like I inferred from you post, and I agree with, each set of demographics may affect the performance outcome differently, across the spectrum of attributes and metrics that need to be considered.
And you're right about the oversubscription phenomenon, as well, but I think that it will be more problematic in the contention type of devices than it will in the switched ones. The contention types which rely solely on Layer 3 routing protocols are, however, less expensive, they are easier to administer, and they run on their own like a top, without having to define and administer ATM virtual paths (VPs) and virtual circuits (VCs) on a per user basis (often multiple VCs per user, at that).
But the potential down side in the exclusively-Layer 3 model is the contention itself, and all that that implies, going forward.
Granted, this space is in the midst of having a newly defined set of cultural benchmarks and baselines to measure against, and a lot of this has to do with marketechture, rather than architecture. It reminds me of the Sprint "Zero CIR" offering in their frame relay service, which meant that Sprint would guarantee a certain level of service on their frame relay access lines and ports, without the customer having to pay a premium for a committed information rate, hence the term Zero CIR.
The frame relay marketing arena was, and continues to be, an interesting study from a number of perspectives. Do you suppose we'll see something similar to CIRs and guaranteed data (throughput) rates in the local loop? I do, but don't ask me when.
If not, this whole discussion is moot, since there will be no method of vindicating or disproving any of the carriers' claims... or the users gripes, for that matter.
And I suspect that a lot will center on motivation as well. Why a carrier would either overbuild their provisions, or underbuild it, is a matter of the intended effect they want from the realized performance levels. Are they wholesaling it or retailing it? Is there competition in the area? Can they do better in bundling other services if the performance and retention of the DSL service remains high? Is there something else that they'd rather be selling? Is the churn rate too high in a given area? The answers to a lot of questions go into their ultimate motivation.
In a highly competitive area where high-end users are extremely performance sensitive, I suspect the carrier would be more motivated to maintain a higher level of performance, especially if there were cable modem and wireless competitors breathing down its neck. And vice versa in service areas with lower visibility (typically referred to as under-served areas).
Which all leads me to a question that I've thought about at times, and that is how does the PSC come to measuring such performance within the frame work that they must, if they are going to gage the carrier's performance in the delivery of next generation technologies such as these?
In dial tone there are means of measuring all kinds of parameters, everything from dial tone delays, % call completions on first attempt, number of missed appointment dates, etc. Do you suppose that the carriers will be forced into providing the PSCs with hot links into their DSL Bandwidth Management, and other network management, systems? Enough for now, someone else take over here for a while. Take it, Bill!
Frank C. |
