DISQUS

lippard · 2 years ago

A good place to start on peering is Geoff Huston's "Interconnection, Peering, and Settlements" and Pio Baake and Thorsten Wichmann's "On the Economics of Internet Peering" (PDF).

lippard · 2 years ago

Whoops, misplaced a tag there, but both links should work.

False Data · 2 years ago

If you're looking at packet shaping and prioritization as a handle on the smart network/dumb network debate, you might also consider a third technique which proved effective in practice: have your company connect to multiple service providers and steer outbound traffic down one link or the other. It takes advantage of the fact that traffic tends to be asymmetric, with most of outbound from the company. (Picture someone like YouTube using this stuff.) Routescience used to have some stuff called Adaptive Network Software that did that. (You can find a quick overview here, but it helps if you speak NANOG :-) . They've since been acquired by Avaya but might be willing to furnish whitepapers. I think Sockeye had something similar, too. A good general lesson to draw is that getting good performance across a wide range of applications without embedding intelligence in the network can be a difficult technical challenge.

For other references, I assume you've already gone trundling through the cites in Rob Frieden's Network Neutrality or Bias paper?

Richard Bennett · 2 years ago

The patron saint of network diversity is Larry Roberts, one of the fathers of the Internet. His paper Changing the internet to support real-time content supply from a large fraction of broadband residential users is very instructive.

Also, anything you can find on 802.11e should be helpful. This amendment to the 802.11 standard incorporates priority-based QoS as well as parameterized QoS, and it's very widely used.

On the economics, your best source is Chris Yoo.

Martin Geddes · 2 years ago

You might find my article on Paris Metro Pricing relevant: http://www.telco2.net/blog/2007/01/rethinking_q...

It's a "dumb QoS for dump pipes" method that preserves end-to-end.

Richard Bennett · 2 years ago

Paris Metro Pricing on a packet network has some interesting side effects, and it's already quite popular; overlay networks such as WebEx use it, for example.

Preservation of end-to-end, if it's important, isn't really affected by a QoS menu, however. End user packet streams choose from several transport options rather than being forced into one class all the time.

MG · 2 years ago

Whatever you do, model the scenario correctly: i.e., the broadband providers are gatekeepers for the content, and are not content providers themselves. So, it is a two-way market, which most studies ignore.

False Data · 2 years ago

OK, I had a chance to do a lot of reading about the economics over the weekend. I've published an annotated bibliography.

Tim Lee · 2 years ago

Thanks, that's helpful!