Send As SMS

Monday, June 12, 2006

"Local" Internet

Cringely proposes an interesting idea: PBS stations could set up truly local delivery of their content over IP by co-locating content servers with local access providers. The PBS stations are ideally placed to service local markets, the transit bandwidth costs could be shaved to [near] zero, the access providers themselves could almost certainly profitably "charge" only an advertising revenue share and, of course, viewers could gain the ability to watch not only live TV over IP, but far more usefully in most cases, previously-aired programmes on demand.

At first blush this seems a little off-the-wall (as many of Cringely's ideas do :-)) but upon some reflection this one's starting to grow on me. The first thing that occured to me was that what he's proposing is essentially a simplification of an existing technology, specifically Akamai-style CDNs. This rang a bell; Clayton Christensen points out in Innovator's Dilemma that, time and again, there has been enormous mileage to be had in repackaging a simplified version of an existing technology and selling it in a new market far more cheaply to far more customers.

(The Dilemma is about the particular problems for an encumbent to do this because, often, the cheaper repackaging will in fact be adequate for some of the encumbent's existing customers, so producing and marketing the simpler, cheaper product will cannibalise some of the existing revenue stream. For Cringely's proposal, it does not matter whether the implementor is Akamai, a startup, or the PBS stations themselves.)

Upon further reflection, some refinements present themselves:

  • It's not neccessary to co-locate gear at access provider's premises, at least not in all cases, instead it would probably be enough for a station to place a couple of servers at each of the two or three nearest NAPs and offer an advertising revenue share to each access provider who has a presence at one or more of those NAPs, in exchange for the access required to deliver to that access provider's local customers. The cost-per-access-provider then drops to the cost of an Ethernet cable! In fact, given that all that is required is half a dozen servers plus rackspace, the total infrastructure outlay is a few thousand dollars up front plus a couple of hundred per month.
  • This approach gets content delivery through (essentially) all local access providers without having to pay a cent for transit. It may even make sense to not provide a reduced service for rest-of-Internet and, thereby, not have to buy transit for delivery at all. This is certainly in line with the simplification mindset and may make the difference between making a pilot feasible or not.
  • Given that the stations are serving local markets, it's not even neccessary to pay transit costs to get content from the studio to the servers, they need merely have access circuits at the studio provided by one or more participating access providers.
A problem also presents itself, although maybe this is a blessing in disguise. I suspect that most access providers will cover larger territories than most PBS stations. This could mean a need for some costly engineering to keep the PBS traffic off an access provider's internal trunk circuits ("this service not available in your area"), or it could mean that PBS stations find themselves with marginally greater coverage, although even this might lead to turf wars. I don't know enough to guess the likely outcome here.

Anyway, it's an intriguing idea.

UPDATE 22-Jun-2006: During a private exhange on this with Cringely he ponted out that to me, of course, contention ratios which are implemented far from the NAP (out in the local BT exchanges in the UK) would render placement at the NAP a little futile; the video server(s) would need to sit at the local exchange(s) to be able to make "cheap" local deliveries.