[ NNSquad ] Re: NYT: Differing views on Time Warner'sBandwidth Cap Experiment

[Lauren Weinstein <lauren@vortex.com>]

A significant problem is that since by and large the network
topologies involved are not publicly known and not regulated in
detail, all such analysis at that level is essentially a guess.

We don't really know how much of any bandwidth contraints are due to
local capacity issues, upstream links between the cable system
"intranet" and the public Internet, or other parameters.

We do know that the broadband provider in this kind of case holds
all of the cards, and on modern cable systems, for every chunk of
frequency bandwidth allocated to their own media servers, or to HD
video channels, or to switched video channels, or whatever along
those lines, there's in theory and often in practice corresponding
less available for the Internet per se.

And even within the Internet frequency allocations, the cable company
has significant control over performance parameters and
subscriber/throughput tradeoffs that directly impact the customer
experience and the usefulness of outside Internet services to those
subscribers (see: "Characterizing Residential Broadband Networks" -
http://www.imconf.net/imc-2007/papers/imc137.pdf ).

Since we don't *really* know what's going on in these regards, it's
impossible to reliably determine if these broadband ISPs are
behaving in a fair manner or not at any given time.  The fact that
they tend to declare so much of this information to be proprietary
helps to ensure the outside world's ignorance and prolong the
guessing game.

Of course, trying to gather metrics to help lift this information fog 
is part of NNSquad's main purpose.

--Lauren--
NNSquad Moderator

  - - -

> Actually, this is NOT quite the same, if/when such a system is
> deployed, there will probably be a singificantly lower transport cost
> for the cable co's offering.  And I'm willing to bet this is the sort
> of argument you'll see:
> 
> Here's why:
> 
> For the bits from Competitor, they are coming over the commodity
> internet link.  Often, its not the local network thats oversubscribed,
> but the upstream link.  Likewise, there is an additional cost with the
> upstream link, which can be significantly more than the cost of the
> local loop for bits actually sent.  (Local loop is cost of building
> the infrastructure, while upstream bandwidth is a continuous cost).
> 
> With the cable company's service, its coming from the local-office
> disk array, perhaps even a cache located in the distribution system
> close to the endpoint, which takes special advantage that it is
> streaming video.  Since it is video streaming, a 1 GB/hr video source
> requires serving only .27 MB/s.  Thus a small, ~$500, passively cooled
> system with a single 1 TB disk could cache 1000 hours and serve 100+
> simultaneous customers.  Play multicast games and the caching could be
> nice and efficient.
> 
> So, actually, the transport cost for the cable company for its own
> service IS likely to be less than transferring bits for the
> competitor.  So why shouldn't billing reflect this?
> 
> The interesting question is "Will they bill for bits from the Akamai
> CDN box?"  Since those bits aren't coming over the commodity Internet
> either.
> 
> 
> Also, at $.10-$.18/GB overage charging (what Amazon charges for their
> service, which would probably be a good guideline for network costs
> with a nice profit to the ISP ), we aren't talking a very big penalty,
> perhaps a dime or two, for transferring from the third party video
> source.  So the "competitive advantage on pricing" on these $5
> pay-per-view/"rental" movies is 5%.
> 
> 
> On Jan 18, 2008 2:22 PM, Lauren Weinstein <lauren@vortex.com> wrote:
> >
> > Now we get squarely into core Network Neutrality issues.  For
> > example, when you order a PPV movie from your cable company, it's
> > just bytes being pumped down the line on the same physical cable as
> > your cable Internet connection.  However, you can be sure that the
> > bytes used to send you that film from your cable company servers
> > won't be counted against your monthly allocation of Internet data.
> >
> > So, if you had a choice of paying N dollars to your cable company to
> > get a movie without affecting your Internet usage totals, or
> > alternatively paying the same amount to a cable company competitor
> > for the film, but using up a significant fraction of your monthly
> > Internet allocation in the process, which would most people
> > presumably choose?
> >