Mike:
When I designed switch architectures, there were Three Things To Remember (3T2R as we say in telecom) that we considered
important in costing a switch:
Number 1: line circuit cost
Number 2: line circuit cost
Number 3: line circuit cost.
This is an old joke, of course, that I adapted from real estate's "location location location" , but it is quite true. Except today, when
we say line circuit, we tend to mean "a lot of DSP resources and a bunch of glue logic chips that do wonderful things between analog
and digital or digital and digital for voice and data stuff for a single subscriber or for a single session'. Call it "line circuit" for now, as
any acronym describing that would be far too long.
Now I wouldn't invest in just ANY DSL company. But when you say DSL, it is not xDSL or *DSL. There are two types::
1. There is HDSL
2. There is all the rest DSL, which I will call !HDSL, to use the PERL program language syntax; that's "!" for "not", thus !H is
anything except H
Now HDSL is one-and-only-one thing: a T1 replacement strategy. That is, you take a twisted pair copper, and make it look like a
T1 at either end. This technology will succeed, since most businesses are sub-T1, and even large business have legacy T1
equipment and 24 channels will handle the voice and data needs of 100 people. Going to 10G ethernet using point to point fiber
protocols is strategy that probably won't appeal to most small business (in the short term) who are more worried about getting their
phone bill from $.1 to $.03 per minute then getting Napster to the workers' desks, for example.
Ok, in the future, with video conferencing and distance learning, there may be more demand requiring fiber to the office, but HDSL
will be a winner in the short run, I believe. ADCT was quite clever in buying PAIR at a real bargain price, and will dominate this
business with 55% market share- see corporate-ir.net. Paradyne
is also in this business, and recently announced a shortfall and got whacked as you noted, although that may be due to other reasons
(such as !HDSL product failure); I don't know. Perhaps we are reaching market saturation, or fiber to the office is winning, but I
don't think so. Yet.
Now all the other DSLs are basically local access and have a problem called the DSLAM. The DSLAM is a box, external to the
switch, that splits voice off from data. The voice goes to the switch, as analog or possibly T1's with TR303 signaling, and the data
goes out over 100Mbps ethernet, or whatever (they are most happy with FrameRealy, I think). Whichever solution you try, you
have a problem- 2 line circuits, at least one of which is analog. Remember the 3 Things To Remember (3T2R) above? You've
perhaps doubled the cost of a line switch. Ouch!
So the obvious thing to do, and I say this as a former switch architect, is to move the DSLAM line circuit and its nice little DSL into
a line circuit card that fits into the backplane of the switch, so you have only one line circuit cost point. I note that WESTEL was
working with LUCENT to do this- although from some press releases it seemed copper mountain is also working to do it, and
LUCENT has a track record of doing things in house as well. In any case, this is the smart, or at least cheap way to add !HDSL to
your switch, and these !HDSL companies may do ok, except there is strong cost pressure so they will have low margins (as you
point out in your ORKIT note).
On the other hand, !HDSL only goes so far from the CO. And in other cases there is a 'box' (you probably have seen those
padlocked gray boxes scattered about near a housing development) which collect a bunch of analog lines and covert them to T1 (or
now optical) and remote them back to the CO. Well, there is a market to put a 'harsh environment' DLSAM card in these boxes, or
build a new !hDSL box and deploy DSL that way. But again, you will still double the line cost!
So, in conclusion, !HDSL is a razor edge slim margin business, unless you build it into the switch, where it is, well, a razor edge slim
margin business still. HDSL is a wonderful business with ripe margins for a few years, but does not have a long long term future.
Fiber to the office will likely take over (this is starting to happen even with single T1 business, but for strange reasons), running we
don't know yet (although we all have correct and firmly held opinions!).
Now you note: Overall what this... [recent problem in Orkit and Paradyne]... tells me is the difficulty a equipment company
faces trying to compete for business in the cheap, cheap, incumbent twisted copper pair
world. That's all the more reason I think the twisted pair world is loathe to give up any legacy hardware and/or networks
until it is absolutely, 100%, forced to.
Thus, I agree with you, but at greater length. We will keep legacy equipment inside the offices. However, since HDSL is a way to
exend the life of legacy hardware but reduce the transport/access costs, I think it is a winner
Just to cause further trouble, I would note for those of us who believe in VOIP that, hey, VOIP requires two line circuits (or rather,
a H.xxx/RTP/UDP/IP/DSP and a T1/TR303/DSP circuit. But that's different, isn't it?
justone opinion |
