[101912] in North American Network Operators' Group
Re: Lessons from the AU model (was: An Attempt at Economically Rational Pricing: Time Warner Trial)
daemon@ATHENA.MIT.EDU (Andy Davidson)
Sun Jan 20 18:41:43 2008
Cc: nanog@merit.edu
From: Andy Davidson <andy@nosignal.org>
To: Andrew Odlyzko <odlyzko@dtc.umn.edu>
In-Reply-To: <20080120163721.38F183C0E2@dl1.dtc.umn.edu>
Date: Sun, 20 Jan 2008 23:40:48 +0000
Errors-To: owner-nanog@merit.edu
Hi,
On 20 Jan 2008, at 16:37, Andrew Odlyzko wrote:
> The more sensible end of town pays about $80 per month for about 40 =20=
> Gbytes of quota, give or take, depending on the ISP. After that =20
> they get shaped to 64 kbps unless they want to pay more for more =20
> quota.
I replied offlist to Andrew with some ideas, but I have been thinking =20=
more about the econometric model of Australian connectivity, and how =20
interesting it is.
> When transit is costing $250 per megabit per month, there aren't =20
> many other options.
[...]
>> On Sun Jan 20, Matthew Moyle-Croft wrote:
> Having a cap and slowing down afterward (64kbps or 128kbps are =20
> typical) is what worked here in Oz.
The grass is always greener of course, but when I think about why ISPs =20=
in the UK have to cap fairly aggressively (bear in mind I pay the =20
figure Andrew cited is typical for internet access in AU, and have a =20
smaller quota !), there are aspects of the Australian problem that I =20
am envious of.
Australia has a relatively small population (c. 20m) which would act =20
as a small ceiling for demand, but the vast majority of the population =20=
live in relatively close proximity. Density is highest in coastal =20
regions which makes it ideal for fiber landing ! I will trust =20
Andrew's numbers of $80 for a 40GB cap.
The UK has a population of c. 60m, and population density is high (12k =20=
people per square mile in London). I'm more likely to pay less than =20
$10/Mbit for global transit. The small country and concentration of =20
POPs in key metropolitan areas makes interconnection cheap. Exchange =20=
membership and participation is hugely popular (>600 networks peer in =20=
London, 415 peer exclusively in the UK). And yet I pay US$70 to my =20
ISP for residential connectivity and my cap is 30GB. Why is this ? =20
Thanks to the pricing model imposed on last-mile connectivity imposed =20=
by the incumbent, it costs an ISP US$471/Mbit to send data to my =20
customer[1]. Maybe the same data that's just come all the way from Oz =20=
through my transit for US$10.
It used to be the case that global transit was very expensive wherever =20=
deployed in the world. As pricing fell, this fueled innovation and =20
created demand for connectivity at every level - domestic, data =20
centre, enterprise, carrier ... The price of connectivity to =20
Australia is likely to fall, because as the opportunity to sell =20
connectivity increases, so should the number of fibers running to =20
Australia (if we were all in the same room, this is the point that Rod =20=
would wave his arms, leap out of his seat, and announce that he's =20
already half the way there and would get there first, so noone else =20
needs to try :-) ). And as economies way off to the west of the =20
country grow and stabilise, then the options for sea routes to the =20
country will grow.
In the mean time, that $250 figure is a market price. Attempt to =20
modify the market conditions will change the price. But perhaps there =20=
are commercial activities that could stimulate demand for consumer IP =20=
services - here's some ideas for thought
- What's my traffic to south Asia and the other apnic regions ? Can =20=
I save some money buying *partial* routes from a large player in this =20=
region. Or is the problem that actually it's the transport to =20
*anywhere* out of Australia ?
- Am I peering widely enough ? Should I actually be stuffing a =20
switch under the floor in my employer's suite and letting my buddies =20
plug in ? Peeringdb knows about eight exchanges in a developed =20
economy of 20 million people. We have more than eight in single =20
cities of Europe.
- So transit pricing sucks. But that's one of my costs as an =20
operator. What's the pricing of a footprint in carrier hotels ? Are =20=
there enough carrier hotels ? How much am I paying for power ? If =20
real estate and power is cheap in AU, then shouldn't content network =20
operators in places where power and space is expensive already be =20
planning how to pop up in Australia ?
- What about local content ? Why is so much traffic leaving the =20
country ? Does someone need to be extremely plucky and offer bargain =20=
basement content hosting in the continent ? If AU entrepreneurs are =20
ignoring the online channel for retail and entertainment, then who =20
wants to jump on a plane and turn this situation around with me ? :-)
- OK, how about we proxy certain types of content unless our users =20
opt out. Any cache hits don't contribute to their 40GB monthly =20
download. If transit is the problem, then offer financial incentives =20=
to your users to help you not pay for it. If you're peering IP, why =20
not start "peering" your top cache hits between providers too ?
- P2P is probably a problem for AU networks. Contrary to most =20
policy makers, there are services which use p2p as a transport, that =20
don't involve the distribution of copyright material without consent. =20=
The next generation services that use p2p as a transport, e.g. Joost, =20=
have said to be trying to build proximity awareness into their p2p =20
implementation. Peering widely will help here. As for the file =20
sharers, then (I really, really hate to say it) but can you just make =20=
sure you're picking up the seedier parts of usenet binaries over =20
peering instead, and hope people use that ? Sad to think in these =20
terms, but if we're being pragmatic ...
I'd love to hear the opinions of AU operators on these issues, and =20
think that there's lessons for everyone - if AU operators can show us =20=
how they deploy more cost effective connectivity products, then there =20=
are some regional ISPs in the rest of the world who would also benefit.
Andy
[1] =A31,758,693 ($3.5m) PA for a 622Mbit BT Central, (so in bandwidth =20=
terms, equates to $471/Mbit per month - if the central is maxxed out) =20=
- I posted this to Nanog in October.=
