Mike,
>So even if every single neighbor of mine, connected to my local CO, decided to subscribe to 24x7, xDSL service, there would be no problem with the local CO not being able to place voice calls? Right? And it's all because of the splitter, right? <
In those varieties that employ splitters, you are correct. But those are of a dwindling variety.
OTOH, I would not universally characterize the data portions of those DSL lines as switched services. They are, instead, merely access lines which will support a variety of different types of protocols and modalities. And the assurance that you have of "getting through" and at what data rate, may depend on the type of DSLAM the provider is using, how the provider has "sized" the network, and what the level of contention is in the pool of users you are vying against. To a great degree, the concentrator version of DSLAMs resembles the traffic dynamics of Cable Modems on a shared segment.
We come back to the discussion that Denver Techie touched on earlier today, and my previous posts concerning multiplexed (switched/deterministic) DSL versus concentrated (routed/best effort) DSL.
In a DSLAM of the multiplexed type, you are either go or no go opn an instantaneous basis, since the switching on the backplane is deterministic. I'm not sure, however, how many of these ATM-based devices actually use back pressure and congestion controls. Maybe someone can answer that question?
On this type of DSLAM which employs layer 2 switching on the backplane, there may be contention, and if you are restricted, it is absolute for that moment of the attempt. Or, more likely, you are buffered and eventually "cut through" with a slight delay. But you are not sharing a common collision domain on the backplane at any time, like you would be on an Ethernet domain, at any point in time.
This should only be a problem less than ~0.2% (<?>) of the time if the carrier has sized the plant provisions correctly, which means they've taken all of the traffic dynamics into account, correctly. The larger carriers will normally do this. The shared tenant building owners, and the fly by night ISPs? Don't bet on it. Believe me on this: Don't bet on it!
In a concentrator type which employs ethernet-like contentioin and routing, on the other hand, there is an intervening collision domain between the loop side of the box and the router side, and perhaps a switched ethernet uplink to the router port(s). In this situation you have both an increased risk of security leakage, and the possibility of aborted attempts, due to the nondeterministic nature of the beast. In other words, it becomes one of best effort, and when you stall in the box, your TCP takes over and requests a retransmission, further aggravating the situation.
By the way, I prematurely hit the submit on this about fifteen minutes ago, and my time for editing is running out, so.... |
