http://arstechnica.com/news.ars/post/20060331-6498.html
"In the foreseeable future, having a 15 Mbps Internet capability is irrelevant because the backbone doesn't transport at those speeds," he told the conference attendees. Stephenson said that AT&T's field tests have shown "no discernable difference" between AT&T's 1.5 Mbps service and Comcast's 6 Mbps because the problem is not in the last mile but in the backbone."
Is this something held generally true in the US, or is it just pointed hair-talk? Sounds like "nobody should need more than 640kb of memory" all over again.
I can definately see a difference between 2 meg, 8 meg and even faster, even when web browsing, especially transferring large pictures when running gallery or alike. When I load www.cnn.com with 130ms latency I get over 1 megabit/s and that's transatlantic with a lot of small objects to fetch. Most major newspapers here in Sweden will load at 5-10 megabit/s for me, and downloading streaming content (www.youtube.com) will easily download at 10-20 megabit/s if bw is available. flickr.com around a couple of megabits/s. (all measured with task-manager in XP, very scientific :P)
I can relate to there being a sweetspot around 1.5-3 megs/s when larger speed doesn't really give you a whole lot of more experience with webbrowsing, but the more people will start to use services like youtube.com, the more bw they will need at their local pipe and of course backbone should be non-blocking or close to it...
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Mikael Abrahamsson wrote:
AT&T: 15 Mbps Internet connections “irrelevant” | Ars Technica
"In the foreseeable future, having a 15 Mbps Internet capability is
irrelevant because the backbone doesn't transport at those speeds," he
told the conference attendees. Stephenson said that AT&T's field tests
have shown "no discernable difference" between AT&T's 1.5 Mbps service
and Comcast's 6 Mbps because the problem is not in the last mile but in
the backbone."
Is this something held generally true in the US, or is it just pointed
hair-talk? Sounds like "nobody should need more than 640kb of memory"
all over again.
I can definately see a difference between 2 meg, 8 meg and even faster,
even when web browsing, especially transferring large pictures when
running gallery or alike. When I load www.cnn.com with 130ms latency I
get over 1 megabit/s and that's transatlantic with a lot of small
objects to fetch. Most major newspapers here in Sweden will load at 5-10
megabit/s for me, and downloading streaming content (www.youtube.com)
will easily download at 10-20 megabit/s if bw is available. flickr.com
around a couple of megabits/s. (all measured with task-manager in XP,
very scientific :P)
I can relate to there being a sweetspot around 1.5-3 megs/s when larger
speed doesn't really give you a whole lot of more experience with
webbrowsing, but the more people will start to use services like
youtube.com, the more bw they will need at their local pipe and of
course backbone should be non-blocking or close to it...
Sounds like FUD to me...
Perhaps trying to downplay the push to FIOS???
- --
Date: Sat, 1 Apr 2006 08:34:36 +0200 (CEST)
From: Mikael Abrahamsson
AT&T: 15 Mbps Internet connections “irrelevant” | Ars Technica
"In the foreseeable future, having a 15 Mbps Internet capability is
[ snip ]
Is this something held generally true in the US, or is it just pointed
hair-talk? Sounds like "nobody should need more than 640kb of memory" all
over again.
I think the Comcast and "cheaper cable plant" references answer your
question. With "new AT&T" adverts, political lobbying, selling retail
DSL below loop/backhaul-only, and consolidation costs, how much money is
left over for last-mile upgrades?
Call me cynical. I just seem to recall AT&T ads in US news magazines
bragging about backbone size _and_ the large portion of Internet traffic
they [supposedly] carry. (I say "supposedly" because claims might be
technically true, but misleading, when traffic passes over AT&T _lines_
via other providers' IP networks. Shades of UUNet and Sprint[link] from
years gone by, anyone?)
So... uh... assuming all three claims -- "backbone is bottleneck", "we
have big backbone capacity", and "we carry big chunks of Internet
traffic" -- are true... I'm puzzling over what appears a bit
paradoxical.
The IPTV reference is also amusing. Let's assume a channel can be
encoded at 1.0 Mbps -- roughly a 1.5 hr show on a CD-ROM. I don't see
two simultaneous programs, Internet traffic, and telephone fitting on a
DSL connection.
Perhaps the real question is which regulatory agency, or shareholders,
needed to hear what the article said. 
Eddy
Google for: telecommunications bill 2006
Eddy
You can listen to Randall Stephenson's presentation at the BoA conference
at the site:
http://www.veracast.com/webcasts/bas/media06/id98101155.cfm
This particular topic is in the Q&A towards the end of the talk.
It was a financial analyst conference, so the technical language was
probably a bit loose. AT&T has an OC192+ backbone, so obviously it
wasn't a technical answer. At other conferences, other speakers have
publically said they are also looking at bonding pairs to get even
greater link speeds (40-100Mbps), not to mention other dedicated
Internet access products with even faster link speeds. You have second
phone lines, why not second DSL lines for people who feel the need for
speed? Likewise cable modems (DOCSIS3.0) are adding channle bonding for
higher access link speeds.
But I think Mr. Stephenson's point was a network bottleneck is not always
based on the access link speed some ISPs put in their advertising. Just go
to any ISP user forum and you will see long threads complaining they can
only download X Mbps from site Y in city Z. The bottleneck may be the
remove server, a peering interconnect, a backbone link, a city router,
etc. On the other hand, its not a good idea to generalize because other
users in other cities may get better performance from other sites.
There are also differences in how people use the network. Power
users and gamers are looking for any edge they can get. Casual users
may be more price sensitive and may not perceive enough of a difference
between 6Mbps and 16Mbps for what they do. If you consider it from a
marketing point of view rather than a technical point of view, if you
are a mass marketer where do you find the biggest target markets?
Wal-Mart targets a specific price point and target market and is very
successful even though it doesn't sell ultra high-end goods.
That's not to say things are static, and will never change. If you
listen to Stephenson's presentation, he says access link speeds will
increase, as well as the backbone capacity will increase. For financial
analysts, the foreseeable future is the next quarter's financial
results. Next year is long term. Two years is an eternity.
[snip]
But I think Mr. Stephenson's point was a network bottleneck is not always
based on the access link speed some ISPs put in their advertising. Just go
to any ISP user forum and you will see long threads complaining they can
only download X Mbps from site Y in city Z. The bottleneck may be the
remove server, a peering interconnect, a backbone link, a city router,
etc. On the other hand, its not a good idea to generalize because other
users in other cities may get better performance from other sites.
Bing. With the dayjob hat of a high speed access provider, we get no
end of issues on everything from end-users not "getting their full speed"
due to loacl or remote using 802.11b or lousy TCP implementations (untuned
win98) or the like. These occur just as often as remote hosts on a DS1
or swamped 10M access, etc.
That's not to say things are static, and will never change. If you
listen to Stephenson's presentation, he says access link speeds will
increase, as well as the backbone capacity will increase. For financial
analysts, the foreseeable future is the next quarter's financial
results. Next year is long term. Two years is an eternity.
This also bears repeating. The investment community's timescale is
dramatically different than that of the decisionmakers for capital
spending in most companies.
Joe
Since AT&T provides nearly all of the transit bandwidth to Comcast in New England, this thread says to me, more or less, "those folks at Comcast claim speeds they can't deliver, because the backbone they use -- which happens to be AT&T's -- is too congested to deliver those speeds anyway." Or something like that.
Yes, clearly I'm poking fun at AT&T here. Large providers who want to play in both the wholesale and retail space really should think about how their marketing in one area affects their claims in another. That's a non-marketeer's view, clearly.
If AT&T is really claiming that their backbone has less than 15 Mbps capacity (which
is how "the backbone doesn't transport at those speeds" reads in plain English), this is either
- an April Fools joke or
- pitiful.
Regards
Marshall Eubanks
If AT&T is really claiming that their backbone has less than 15 Mbps capacity (which is how "the backbone doesn't transport at those speeds" reads in plain English), this is either
Maybe they meant that the typical end-user windows IP stack has small enough TCP windows that when you take into account typical latency across the internet, those users just can't utilize their high bandwidth links due to the bandwidth * delay product.
- an April Fools joke or
- pitiful.
Could be either. Did you happen to catch the woman from Verizon at the last NANOG who was sure parts of New Orleans were 2 miles below sea level? Maybe that was a really early AFJ.
The majority of U.S.-based IP TV deployments are not using MPEG-4, in fact,
you would be hard-pressed to find an MPEG-4 capable STB working with
middleware.
SD MPEG-2 runs around ~4 Mbps today and HD MPEG-2 is ~19 Mbps. With ADSL2+
you can get up to 24 Mbps per home on very short loops, but if you look at
the loop length/rate graphs, you'll see that even with VDSL2 only the very
short loops will have sufficient capacity for multiple HD streams. FTTP/H
is inevitable.
Frank
Date: Sat, 1 Apr 2006 14:02:13 -0500 (EST)
From: Jon Lewis
Maybe they meant that the typical end-user windows IP stack has small enough
TCP windows that when you take into account typical latency across the
internet, those users just can't utilize their high bandwidth links due to
the bandwidth * delay product.
I wondered the same at first, but that's hardly "the backbone". And TCP
windowing affects single TCP streams... with bittorrent and similar,
people have found workarounds.
Eddy
Maybe it's the lost city of Atlantis or maybe she was confused about meters vs. miles. She does work for Verizon...
-Robert
btw-We all know Atlantis is really in the Pegasus galaxy now and not in NOL. 
Tellurian Networks - The Ultimate Internet Connection
http://www.tellurian.com | 888-TELLURIAN | 973-300-9211
"Well done is better than well said." - Benjamin Franklin
No the problem is at AT&T's congested peering edge. 
Date: Sat, 1 Apr 2006 13:26:51 -0600
From: Frank Bulk
The majority of U.S.-based IP TV deployments are not using MPEG-4, in fact,
you would be hard-pressed to find an MPEG-4 capable STB working with
middleware.
*nod*
Again, I don't see how AT&T can claim "DSL is fast enough" in one
breath, then turn around and say they're ready to deliver IPTV.
I'm curious how program content is currently stored. (Note that I'm
totally ignoring live broadcast.) If MPEG-2, I'd guess conversion to
MPEG-4 might produce less-than-desirable image quality.
Eddy
The majority of U.S.-based IP TV deployments are not using MPEG-4
Agreed. However, I'd say that any IPTV provider currently using MPEG2 would
be planning a migration to MPEG4/H.264 - half the bandwidth means double the
channels.
in fact,
you would be hard-pressed to find an MPEG-4 capable STB working with
middleware.
I disagree. There are several MPEG4 capable STB available now, and they all
have support of middleware vendors.
SD MPEG-2 runs around ~4 Mbps today and HD MPEG-2 is ~19 Mbps. With ADSL2+
you can get up to 24 Mbps per home on very short loops, but if you look at
the loop length/rate graphs, you'll see that even with VDSL2 only the very
short loops will have sufficient capacity for multiple HD streams. FTTP/H
is inevitable.
Anyone looking to do HD will be looking at H.264, and looking to bring the
bandwidth requirement down to 8-10Mbps. That is certainly more practical with
ADSL2+ deployments (unless you want more than one STB per DSL).
Simon
(Currently working on an H.264 IPTV deployment)
Whilst MPEG-2 for broadcast purposes will be in the 3-5Mbps range, MPEG-2
for archival/storage will be at a significantly higher bitrate. If you're
storing at high bitrate MPEG-2, the transcoding to MPEG-4 will have much
better results than if you transcoded from broadcast quality MPEG-2.
Simon
Since AT&T provides nearly all of the transit bandwidth to Comcast in
New England, this thread says to me, more or less, "those folks at
Comcast claim speeds they can't deliver, because the backbone they
use -- which happens to be AT&T's -- is too congested to deliver
those speeds anyway." Or something like that.
If you listen to Comcast's presentation to financial analysts, you
know Comcast has already announced plans for its own 40Gbps backbone
using dark fiber leased from Level 3. There are probably a few howlers
in Comcast's presentation too, but network geeks aren't the intended
audience.
Financial analysts probably joke about technologists making presentations
and getting tongue-tied about basic accounting terms too.
Yes, clearly I'm poking fun at AT&T here. Large providers who want to
play in both the wholesale and retail space really should think about
how their marketing in one area affects their claims in another.
That's a non-marketeer's view, clearly.
The next few years should be very interesting, and provide lots
of fodder for pundits everywhere. Comcast buys backbone service from
AT&T. AT&T buys programming for several video channels from Comcast.
They'll both need to exchange phone calls with each other in every city
they sell local phone service.
While its great fun to make fun of pointy-haired bosses everywhere, it
may be more useful to look past the various foot-in-mouth statements
and try to understand what each of them is trying to accomplish. Even
though they are all fierce competitors, they also all do business with
each other.
I archive NTSC video in MPEG-2 at roughly 30 Mbps.
That way, there are no worries about future codecs being too good for the archives.
Regards
Marshall
Hello;
The majority of U.S.-based IP TV deployments are not using MPEG-4
Agreed. However, I'd say that any IPTV provider currently using MPEG2 would
be planning a migration to MPEG4/H.264 - half the bandwidth means double the
channels.
Also, I think that the majority of IP TV deployments right now are not in the US.
in fact,
you would be hard-pressed to find an MPEG-4 capable STB working with
middleware.
I disagree. There are several MPEG4 capable STB available now, and they all
have support of middleware vendors.
In the last IPTV trade show I went to (TVoDSL in Paris in January), I don't recall a single
MPEG STB or IPTV system vendor who wasn't either showing or promising H.264 support.
SD MPEG-2 runs around ~4 Mbps today and HD MPEG-2 is ~19 Mbps. With ADSL2+
you can get up to 24 Mbps per home on very short loops, but if you look at
the loop length/rate graphs, you'll see that even with VDSL2 only the very
short loops will have sufficient capacity for multiple HD streams. FTTP/H
is inevitable.
Anyone looking to do HD will be looking at H.264, and looking to bring the
bandwidth requirement down to 8-10Mbps. That is certainly more practical with
ADSL2+ deployments (unless you want more than one STB per DSL).
Which you would in the US, but maybe not everywhere (yet).
Simon
(Currently working on an H.264 IPTV deployment)
--
Regards
Marshall
Simon:
Our regional head-end is adding MPEG4 in the next 3-6 months, so we're on
the same bandwagon. Unfortunately, we've spent $$$ on MPEG2-only STB. It
looks like we could be transport MPEG2 and MPEG4 around our local transport
rings for a long time. We'll use the MPEG4 for customers who want HD, the
rest will get upgraded as we can afford to.
We have a few customers that have 4 or 5 TVs and want STBs for each one of
them, and besides the fact that we have difficulty getting 20 Mbps on
medium-range loops, we end up installing two modems because our BLC
infrastructure is only configured for three streams. This will hopefully be
resolves in future releases.
Yes, there are quite a few MPEG4-capable STB vendors with lots of middleware
vendors standing behind them, but I challenge you to document one
STB/middleware combination in GA. I haven't seen it. Talk to me in six
months, and it will be a different story.
Frank