[94121] in North American Network Operators' Group
Re: Network end users to pull down 2 gigabytes a day, continuously?
daemon@ATHENA.MIT.EDU (Frank Coluccio)
Wed Jan 10 03:26:38 2007
From: Frank Coluccio <frank@dticonsulting.com>
To: nanog@merit.edu
Reply-To: frank@dticonsulting.com
Date: Wed, 10 Jan 2007 02:25:41 -0600
Errors-To: owner-nanog@merit.edu
Gian wrote:
"From a big picture standpoint, I would say P2P distribution is a non-start=
er,
too many reluctant parties to appease. From a detail standpoint, I would sa=
y P2P
distribution faces too many hurdles in existing network infrastructure to be
justified. Simply reference the discussion of upstream bandwidth caps and y=
ou
will have a wonderful example of those hurdles."
Speaking about upstream hurdles, just out of curiosity (since this is merel=
y a
diversionary discussion at this point;) ... wouldn't peer-to-peer be the L=
EAST
desirable approach for an SP that is launching WiFi nets as its primary fir=
st
mile platform? I note that Earthlink is launching a number of cityscale WiF=
i nets
as we speak, which is why I'm asking. Has this in any way, even subliminall=
y,
been influential in the shaping of your opinions about P2P for content
distribution? I know that it would affect my views, a whole lot, since the
prospects for WiFi's shared upstream capabilities to improve are slim to no=
ne in
the short to intermediate terms. Whereas, CM and FTTx are known to raise th=
eir
down and up offerings periodically, gated only by their usual game of chick=
en
where each watches to see who'll be first.
Frank=20
On Mon Jan 8 22:26 , Gian Constantine sent:
>My contention is simple. The content providers will not allow P2P video as=
a
legal commercial service anytime in the near future. Furthermore, most ISPs=
are
going to side with the content providers on this one. Therefore, discussing=
it at
this point in time is purely academic, or more so, diversionary.Personally,=
I am
not one for throttling high use subscribers. Outside of the fine print, whi=
ch no
one reads, they were sold a service of Xkbps down and Ykbps up. I could not=
care
less how, when, or how often they use it. If you paid for it, burn it up.I =
have
questions as to whether or not P2P video is really a smart distribution met=
hod
for service provider who controls the access medium. Outside of being a ser=
vice
provider, I think the economic model is weak, when there can be little
expectation of a large scale take rate.Ultimately, my answer is: we're not =
there
yet. The infrastructure isn't there. The content providers aren't there. The
market isn't there. The product needs a motivator. This discussion has been
putting the cart before the horse.A lot of big pictures pieces are complete=
ly
overlooked. We fail to question whether or not P2P sharing is a good method=
in
delivering the product. There are a lot of factors which play into this.
Unfortunately, more interest has been paid to the details of this delivery =
method
than has been paid to whether or not the method is even worthwhile.From a b=
ig
picture standpoint, I would say P2P distribution is a non-starter, too many
reluctant parties to appease. From a detail standpoint, I would say P2P
distribution faces too many hurdles in existing network infrastructure to be
justified. Simply reference the discussion of upstream bandwidth caps and y=
ou
will have a wonderful example of those hurdles.
> Gian Anthony ConstantineSenior Network Design EngineerEarthlink, Inc.=20
>On Jan 8, 2007, at 9:49 PM, Thomas Leavitt wrote:So, kind of back to the
original question: what is going to be the reaction of your average service
provider to the presence of an increasing number of people sucking down mas=
sive
amounts of video and spitting it back out again... nothing? throttling all
traffic of a certain type? shutting down customers who exceed certain thres=
holds?
or just throttling their traffic? massive upgrades of internal network hard=
ware?
>Is it your contention that there's no economic model, given the architectu=
re of
current networks, which would would generate enough revenue to offset the c=
ost of
traffic generated by P2P video?
>Thomas
>Gian Constantine wrote: There may have been a disconnect on my part, or at
least, a failure to disclose my position. I am looking at things from a pro=
vider
standpoint, whether as an ISP or a strict video service provider.
>I agree with you. From a consumer standpoint, a trickle or off-peak downlo=
ad
model is the ideal low-impact solution to content delivery. And absolutely,=
a
500GB drive would almost be overkill on space for disposable content encode=
d in
H.264. Excellent SD (480i) content can be achieved at ~1200 to 1500kbps,
resulting in about a 1GB file for a 90 minute title. HD is almost out of the
question for internet download, given good 720p at ~5500kbps, resulting in =
a 30GB
file for a 90 minute title.
>Service providers wishing to provide this service to their customers may s=
ee
some success where they control the access medium (copper loop, coax, FTTH).
Offering such a service to customers outside of this scope would prove very
expensive, and likely, would never see a return on the investment without
extensive peering arrangements. Even then, distribution rights would be very
difficult to attain without very deep pockets and crippling revenue sharing=
. The
studios really dislike the idea of transmission outside of a closed network.
Don't forget. Even the titles you mentioned are still owned by very large
companies interested in squeezing every possible dime from their assets. Th=
ey
would not be cheap to acquire.
>Further, torrent-like distribution is a long long way away from sign off b=
y the
content providers. They see torrents as the number one tool of content pira=
cy.
This is a major reason I see the discussion of tripping upstream usage limi=
ts
through content distribution as moot.
>I am with you on the vision of massive content libraries at the fingertips=
of
all, but I see many roadblocks in the way. And, almost none of them are tec=
hnical
in nature.
>Gian Anthony ConstantineSenior Network Design EngineerEarthlink, Inc.Offic=
e:
404-748-6207Cell: 404-808-4651Internal Ext:
x22007constantinegi@corp.earthlink.net <mailto:constantinegi@corp.earthlink=
.net>
>
>
>On Jan 8, 2007, at 7:51 PM, Bora Akyol wrote:
>=20
>Please see my comments inline:
> -----Original Message-----From: Gian Constantine
[mailto:constantinegi@corp.earthlink.net] Sent: Monday, January 08, 2007 4:=
27
PMTo: Bora AkyolCc: nanog@merit.edu <mailto:nanog@merit.edu>Subject: Re: Ne=
twork
end users to pull down 2 gigabytes a day, continuously?
> <snip>=20
>I would also argue storage and distribution costs are not asymptotically z=
ero
with scale. Well designed SANs are not cheap. Well designed distribution sy=
stems
are not cheap. While price does decrease when scaled upwards, the cost of s=
uch an
operation remains hefty, and increases with additions to the offered content
library and a swelling of demand for this content. I believe the graph beco=
mes
neither asymptotic, nor anywhere near zero.=20
>To the end user, there is no cost to downloading videos when they aresleep=
ing.I
would argue that other than sports (and some news) events, there ispretty m=
uch no
content thatneeds to be real time. What the downloading (possibly 24x7) doe=
s is
to stress the ISP network to its max since the assumptions of
statisticalmultiplexinggoes out the window. Think of a Tivo that downloads
content off theInternet24x7. The user is still paying for only what they pa=
y each
month, and this is"network neutrality 2.0" all over again.
>
> You are correct on the long tail nature of music. But music is not consum=
ed in
a similar manner as TV and movies. Television and movies involve a little m=
ore
commitment and attention. Music is more for the moment and the mood. There =
is an
immediacy with music consumption. Movies and television require a slight de=
gree
more patience from the consumer. The freshness (debatable :-) ) of new rele=
ase
movies and TV can often command the required patience from the consumer. Ol=
der
content rarely has the same pull.=20
>I would argue against your distinction between visual and auditorycontent.=
There
is a lot of content out there that a lot of people watch and thecontentis 2=
0-40+
years old. Think Brady Bunch, Bonanza, or archived games fromNFL,MLB etc. W=
hat
about Smurfs (for those of us with kids)?
>This is only the beginning.
>If I can get a 500GB box and download MP4 content, that's a lot ofessentia=
lly
free storage.
>Coming back to NANOG content, I think video (not streamed but
multi-pathdistributed video) is going to bring the networks down not by
sheerbandwidth alone but by challenging the assumptions behind theengineeri=
ng of
the network. I don't think you need huge SANs per se tostore the content ei=
ther,
since it is multi-source/multi-sink, thereliability is built-in.
>The SPs like Verizon & ATT moving fiber to the home hoping to get in onthe
"value add" action are in for an awakening IMHO.
>Regards
>Boraps. I apologize for the tone of my previous email. That sounded grumpi=
erthan
I usually am.
>
>=20
>=20
>
>-- Thomas Leavitt - thomas@thomasleavitt.org - 831-295-3917 (cell)
>*** Independent Systems and Network Consultant, Santa Cruz, CA ***
><thomas.vcf>
>
