Eric, first I'd like to point out that my observations
concerning this matter are mainly objective ones stemming
from an interest in this new space. I, too, am in the learning
stages, using this board as a sounding board to a great
extent, like many others here.
If I find errors in my assumptions and claims, I'm usually
quick to point them out myself if I find them first. Hopefully I
do that before anyone else has had a chance to, but I do
appreciate when someone more knowledgeable than I in
these matters comments or corrects what I've written.
For a classic example of this type of participatory deference,
I invite you to visit the Last Mile and Silk Road threads. In
the LM where many of these cable topics are discussed
regularly, I will almost always defer to Denver Techie (and
others) for cable infrastructure advice; to WTC (and others)
for DSL and telco plant buildout guidance;
...and Dave Horne who frequents here (and others) when it
comes to cable modem; to Dr. Bernard Levy and Dr.
SteveG (and others) when it comes to broadband wireless;
and to Dr. AHhaha (and few others) in the Silkroad thread,
when it comes to matters of theoretical physics which affect
photonics. Having said that...
"The basic spirit of your post is that additional ISP's
would need to be inserted right at each individual head end
and then assigned a separate frequency channel. That in itself
would seem headache enough to preclude most ISP's from
even undertaking the task."
Concerning the need to adapt to an additional channel: This
would be true if the realizations and decisions were reached
in advance (or found out when it was too late) that a single
downstream 6 MHz channel was insufficient to carry both
ATHM's and AOL's traffic.
My assumptions were that this is already evident and can be
justified using simple arithmetic and some elementary
extrapolations.
Using an additional channel, however, is not an absolute
essential here, since both ISPs can actually coexist and share
one channel. But only one ISP can administer the settings
and controls on the DOCSIS modem at any one time. I
view this as an unlikely occurrence, but this could take
place.
I refer you back to the DOCSIS attributes and
dependencies that I've listed in the preceding post. But if it
were to take place, it would mean that one of he ISPs would
have to take a subordinate role to the other, and we know
how that will work. It's T's plant, and ATHM is part owned
by T... etc.
"The have to run their network to every local head end?
That is essentially duplicating the fiber half of the HFC
network isn't it?"
There are two sides to the head end provisions, the wide
area side (the Internet), and the local loop or drop side (the
HFC).
If you are referring to the wide area Internet side, AOL and
the others don't "have to" run a separate line to every head
end, but they probably will.
Like I stated earlier, they could ride piggy back on the
upstream to the NAPs from T's head ends over ATHM's
infrastructure to some extent through one of many ways of
partitioning the channel on the WAN circuit, and then hop
off to their own ANS backbone (I think they are using), but
this is unlikely.
More than likely, I think, AOL would elect to run their own
T1s, T3s, OC-3s, whatever, on the back ends of the head
end nodes to their larger points of presence, or to the
NAPs proper. Or they could seek full colocation status,
and set up shop in the head end, which would be the equivalent
to what is taking place on the ILEC/CLEC front.
Or something along those lines that would detach them from
ATHM's infrastructure to the fullest extent that they can. I
do not think that AOL relishes the idea of being any more
dependent on ATHM than they have to. But who knows?
This whole affair is in some ways bizarre to begin with,
anything can happen.
"Would this still be true if the functional architecture
mirrored what the simple minded view it as? In other
words that the MSO runs the local connection and the ISP
handles the services like e-mail and content. In that world
view all the ISP's would share whatever bandwidth the
MSO could provide and conceivably charges each ISP for
the bandwidth it uses.After all, we users all share the
bandwidth of the internet itself and accept that as a simple
fact."
Now you are entering some serious architectural territory
that transcends physical and logical delineations. I'll attempt
to genericize to the extent that I can, and indeed neither I nor
anyone else in this still evolving space has all of the answers
to how best this could be achieved.
In simple terms, ATHM's architecture, along with many
other cable modem architectures, is not as straightforwardly
Internet by design as the dial up connection or an ISDN line,
or even the basic DSL service.
Instead of allowing the end point client (the PC and its
internal software provisions) to make application level
adjustments and accommodations, as would be the case
with a dumbed down V.34 modem at 33.6 kb/s simply
handing off streams to the PC, the cable modem itself in
today's HFC design begins to take on some of the upper
layer responsibilities.
These include security, QoS, multimedia, etc., that would
normally be handled by the client.
In other words, the cable modem begins to assume some
burdens normally associated with the client (PC), to the
extent that we cannot discuss it as apples-to-apples
anymore with other end point scenarios which are fed over
dumb facilities.
For the simplest applications which don't require anything
more than IP delivery, perhaps we can. But not when you
start introducing a wide variety of integrated and multimedia
services, and security hooks.
The cable modem has network smarts, is what I'm saying,
and this quality increases the domain that must be
administered during provisioning and the ultimate mediation
of services.
Again, the precepts going into the HFC architectural
decision making processes, which were responsible for what
we have today, were predicated on the assumption that
there would be a single domain administrator, not multiple
ISPs.
The obvious is starting to emerge here, I think, but it is
unspeakable amidst a group of investors who own one of
the ISPs.
Based on the willingness of the principals, or their
acceptance to take strategic write-offs for a greater return
through other arrangements, it's conceivable that the MSO's
optimal role in this scenario, may best be suited to that of
facilities manager and not that of an ISP.
I think that this motif was in the works prior to the advent of
ATHM and RR. It took a turn when the MSOs realized a
means of protecting their enterprises, but these decisions
were made prior to the popularity of the www.
This is not a judgmental assessment on my part, or one that
stems from anything that I could gain, rather, it's an
architectural perspective which is based on the assumption
that the present HFC design will go unchanged.
"Is there a central point to which the MSO brings its data
where the connection to @home is made? Or does @home
currently go out to the headends."
The framework that makes up ATHM's backbone is
depicted on their web site in full color. It's a mesh at the
present time, but their goal is to sonetize it, using DWDM at
the optical layer, with OC-48s above it, to support TCP/IP
in a self healing topology.
Their approach is to use intranet techniques, keeping traffic
in their own domain, with gateways to the outside world (the
larger Internet) to handle off net requirments.
"Is it possible within the routing structure of IP to take a
data stream and split it out to different locations depending
upon who the originating customer is? I suspect this goes to
Ahhaha's original concerns about "tearing the internet apart"
as you put it. Is it possible to meter bandwidth here? Who
is currently responsible for control over the customers
modems, the MSO or @home?"
All of those capabilities are possible, and are being done to
varying degrees with improvements being made daily at the
IETF level, as we speak.
The splitting is taking place through multicasting, metering is
in its infancy and will take place at the higher transport and
session layers, or through the use of ToS indicators, or some
combination of all of the above.
But metering will become a greater issue to contend with
when tiered services come upon us. And control over the
customer's modems is managed through various proprietary
means, in addition to the use of simple network management
protocols, or SNMPs.
But these are administered at the present time by the MSO
in an intranet context, similar to how an enterprise would
treat their own in house IP backbone. The distinction here,
once again, is that the protocols that I have mentioned above
must be reconciled to either one domain or the other (the
open Internet or the intranet), and a duplicity of controls
over those protocols by one private enterprise (ATHM) and
that of another (AOL) cannot be mixed, lest there be some
coming to terms to share administrative roles, which usually
means one delegating responsibilities (read: operational
control) to the other.
"...do you know what % of "upgraded" wires will have
the bandwidth 750Mhz-1Ghz available?"
Sorry, I don't have that information.
"It seems you say this bandwidth is available for
bi-directional use. Can it be used for upstream internet
service or only voice and digital TV as you implied?"
It's intended use is for interactive and multimedia services,
including voice, as well as digital TV delivery... most likely
using MPEG2 as a compression scheme for the latter. The
techniques used for voice are anybody's guess, but I will
assume for the moment that some adaptation of VoIP or
even a purer form of Internet Telephony, will be used in the
future, when it's soup.
Regards, Frank Coluccio |
