Eric, you are a good sport. Thanks for hanging in there with
me both here and over in the ATHM thread, as I exercised
the limits of what if-ing this subject, which is a big part of
what we all do here. I'll try to compile a list of potentially
conflicting constructs between the MSO model and that of
the PSTN/IETF's and get back to you on this at some later
time.
One of your comments in the uplink struck a cord with me,
since it was similar to a question I answered for Harlock
over in MFNX last night. You said,
What amazes me the most is how much of the free cable
spectrum is nominally assigned to non-IP uses. Just assign
50 channels to transport IP and let open standards provide
the rest."
Harlock had asked when I thought we would have Terabit
flows coming into our living rooms. Specifically, he asked,
"...15 years from now when we have streaming terabit connectivity into our
living rooms it will be courtesy of the Muffin. Sorry to be overly enthusiastic but
I've had MFNX since shortly after the IPO so my enthusiasm is understandable...
how many years would you predict to terabits to the living room?"
I replied:
Message 10546578
For the sake of thread continuity here in the Last Mile, in case there are any comments, I've also copied this post, below.
Regards, Frank Coluccio
-----------------------------
To: +Harlock (425 )
From: +Frank A. Coluccio
Saturday, Jul 17 1999 12:33AM ET
Reply # of 427
Harlock,
Some of the points you raise are eye opening, to say the least.
"...15 years from now when we have streaming terabit connectivity into our
living rooms it will be courtesy of the Muffin. Sorry to be overly enthusiastic but
I've had MFNX since shortly after the IPO so my enthusiasm is understandable...
how many years would you predict to terabits to the living room?"
The first part, re Muffin doing the home pulls, I don't know about that. In my estimation
it could happen before then (fiber into the home) but whether or not applications will
be demanding of Terabit capacities is highly questionable. I cannot rationally see it at
this time.
I recently had some difficulty on the MRVC thread a short while back convincing some
of the more enlightened tech investors there that Fortune 100 companies would need
Tb routing and DWDM capacities in two to three years. I think that I finally succeeded
in making my case, but corporate data services are orders of magnitude more
demanding than residential users' requirements. It would be foolhardy of me at this time
to suggest otherwise, although I always leave the door open for enlightenment.
Bringing fiber into the home, however, would create somewhat of a very dramatic
avalanche effect, since every strand is capable of multiple TeraHertz, which translates
into multiple Terabits, variously, more or less, depending on the line coding scheme
used.
The question then becomes, do you actually need this kind of capacity, and if so, how
do you harness this bandwidth in ways which would be useful in the first place?
Interfaces don't exist that can do this yet in an economical way. And certainly, the
entire width of the spectrum on that fiber will not find its way upstream to target sites,
for this would require a backbone that would be roughly 1Tb/s multiplied by the
number of subscribers who would be doing likewise. How many people are there in an
average neighborhood? Multiply that number by 1 Tb/s.
So, although the raw capacity "potential" of the fiber coming into your home at some
point will be higher than a Tb/s, you will likely never get to see anything but a fraction
of that traversing to and from the outside world. Only the part which you can actually
use will be packaged by some means for transport and forwarded to the distant end,
the Internet, or wherever.
But when fiber is brought into the home, the bottleneck will once again shift away from
the last mile and into the network access platform (or the edge and core), somewhere,
but it certainly will not be on the last mile cable or the one that is coming into your
window from the backyard pole, as it is now.
Sometime between now and then (as optical techs improve) we will see a
difference in the manner in which we pick bandwidth off the wire. Today we extract
chunks of bandwidth in time slots, or bytes, derived from various electronic means.
TDM, ATM, Frame, etc.
When fiber manifests in full bloom, we will be able to select bandwidth more elegantly,
and in more abundance through the use of tunable optical detectors. Mini-wavelengths
will constitute what we now regard as "channels" in the electronic sphere, but these
newer information carriers will be far more robust than their purely electronic
ancestors. This is one reason why I will hesitate to put a cap on what I think the new
total bandwidth delivery scheme will comprise.
I simply don't know, but it is conceivable to me that the next level of innovation in this
space will allow for up to some nominal per home capacity of 10 GHz per session
(how many sessions, I don't know) of unchannelized wavelength, on average. This 10
GHz, coincidentally (g), is exactly one tenth the standard 100 GHz channel spacing
allowed in the ITU DWDM model. Perhaps 10 GHz is high? Maybe it's too low.
Maybe it should be for all sessions in the home.
I don't know, since there is no shortage of ways in which the powers that be in this
industry have to inhibit itself, by placing artificial barriers where none need exist. But I
digress...
In the absence of fiber, or of some form of super capacity wireless scheme I've not
come across yet, let's consider what today's cable TV coax brings to you. You already
have the potential of 1 GHz, approximately, coming into your home in the way of the
"black coaxial cable."
Note, I said giga HERTZ, not Gigbits. When modulated this 1 GHz has the potential of
nominally four bits per Hertz, or the equivalence of some 4 Gb/s, at least, of
transmission speed, in total.
Of course, you don't use this kind of capacity for data, instead you only use
somewhere in the vicinity of about 2 Mb/s to 30 Mb/s if you are lucky enough to be a
cable modem user, depending on your service provider's architectural preference and
how you are configured. [I wont discuss dialups here, or ISDN, since they are not
statistically significant for the purposes of this discussion.]
In the preceding example, I only mentioned the 4 Gb/s "potential" against the 2 to 30
Mb/s you are actually using. I did this in order to demonstrate what you have coming
into the home, versus what you are actually using for data at any point in time. The rest
of the pipe is native high frequency analog carrier capacity used for TV signals at this
time, of which you normally only use a single, that is, one (1) 6 MHz channel at a time.
Maybe two or three of them, if you have multiple TV sets running simultaneously, or if
you use PIP.
Worst case, you are actually using less than 50 MHz of capacity if you have three TVs
all viewing different channels, and your cable modem going full throttle, to boot.
That's 50 MHz out of a potential 1000 MHz (or 1GHz), which demonstrates that you
are only using a mere fraction [5%] of the total potential 1 GHz that exists coming into
your home.
But bandwidth in the home is beginning to appear in a growing number of ways.
- Telephone lines (including POTS, ISDN, xDSL Alarm services, meter, etc.),
- wireless devices (pcs, cellular, PDAs, etc.),
- over the air TV and Radio reception,
- entertainment appliances,
- PDA/appliance ancillary devices for synch and transfer,
- and so on.
And then there is another entirely different set of bandwidth consuming devices that are
local to your residence that only attach to the outside world when you say so. These
are Local Area Networks, infra red and wireless synchronization links and data
transfer links, like BlueTooth, and several other emerging standards.
Right now there are actually many professionals and SO/HOs folks who have installed
home servers operating at the 100 Mb/s rate between main and secondary server
devices, and perhaps 10 Mb/s to workstations and peripherals.
In another two years 1 Gb/s copper twisted pair network cards will be available and
cost slightly higher than what the 100s cost today, until they too come down to
sub-$100 dollar costs. In other words, having multiple Gb devices working in the
home will become commonplace, especially when multimedia and other entertainment
based (as well as business type application) uptake kicks in between your own
networked devices.
A Gb/s is only 1/1,000th of a Tb/s, however, and even if all of your devices were set
up for Gb/s operation, you would still be way far from meeting the 1 Tb/s bogie.
The next step up, which the IEEE has already been working on for about a year or
two, is 10 GbE, which some say will be ready for commercial use by 2002-3. At that
time, you're only 1/100th the Tb speed, but this wont happen in the residence, say, until
2003-4, and for early adopters at that, following the same kind of succession as the
100M to 1000M did, and making some allowances for Internet Time acceleration.
Four years from now, in other words, you will have the potential to support nodes that
will be running at 10 Gb/s. Will you have that kind of need by then?
10 Gb/s on a home LAN is the same level of throughput as one gets with an OC-192,
or 192 T-3s, to put it into clearer perspective.
And to put your original question into better perspective, I'd have to ask you and
everyone else here, when do you think that you would require Nineteen Thousand and
Two Hundred (19,200) T3s coming into your living room, which would be the
equivalent of one terabit per second (1 Tb/s)?
Regards, Frank Coluccio
|
