Time Warner Cable’s planned experiment with tiered charging for Internet access has generated a flurry of coverage in the blogsphere, but no new insights (at least that I’ve seen).

The primary problem ISP’s complain about is that 5% of their customers use 90% of the available bandwidth and when they examine this traffic, it’s mostly peer-to-peer file sharing. A reasonable question is how to allow as much of this traffic as possible without increasing an ISP’s variable costs or slowing down their other users.

This may not be as difficult as it appears. Indeed if Internet access was as competitive as mobile telephony, we might already have seen what I’m about to propose - a combination of bundled pricing equivalent to mobile’s “free nights and weekends” and “free on-net calls” with a way to facilitate P2P traffic that leverages exactly these “free” periods.

An ISP’s costs

ISPs have some costs which are relatively fixed and others that are tied to usage. A network is a relatively fixed cost and when it’s not full, the incremental cost of adding traffic is zero! This is the reason mobile operators give away free nights and weekend. They’ve built their mobile network for the peak daytime traffic, so it costs them nothing to run promotions that add incremental traffic at off hours. Peak hours and off hours may be different for an ISP, but the concept is the same. When a data pipe is lightly loaded the ISP’s cost of adding incremental traffic is zero.

On the other hand, some ISP costs are usage based, for example “IP Transit” or more properly, Internet Transit. This is the ISP’s upstream cost to send and receive traffic to/from the rest of the Internet. However, even here, usage-based costs occur at heavy usage. Light usage periods don’t save money. To understand what’s happening, it’s worth a digression on Internet Transit.

Internet Transit

Internet access is monopoly or duopoly or a heavily regulated industry. The middle mile connections from the local network to the Internet backbone may or may not be competitive depending on where you are. But the Internet backbone itself is extremely competitive. If you can get to a major Internet Exchange Point in the US or Europe, there are many providers offering extremely competitive rates for Internet Transit. Typically these services are priced on a megabit per second per month basis (Mbit/s/Month) with lower rates for higher volume commitments. The other key idea is that charges are based on the 95th percentile of all the five minute data rate samples taken during the month. So an ISP can have a few bursts above their typical rate, as long as they represent less than 5% of the sampled intervals.

But this also means there is no extra cost to run at or near the typical rate at all times.

Local traffic

Even more important, if file sharing is done with other computers on the same ISP’s network, then there is no need to pay for Internet Transit at all. The question is how to figure out which potential peers are “on-net” and which are “off-net.”

Sending signals to P2P software

Most P2P file sharing software has relatively little knowledge of locality. Some P2P software practices “prefix awareness,” for example, Joost gives preference to peers in the same /24 IP address block when they are available. But if a major operator provided an automatic way for P2P client software to determine whether a prospective peer’s IP address was currently reachable “for free”, it seems likely the file sharing community would leap on it, and if there’s money to be saved, active file sharers would download the new clients immediately.

A standard way to present such information might be via an extension to the XML-based response codes in one of the whois information exchange proposals, e.g. from ICANN or from APNIC. Also, while what I’m proposing might start as a pricing plan rather like a mobile operator’s “free nights and weekends” and “free on-net calling,” it’s not hard to see extensions where an ISP could offer dynamic access to underused capacity to those programs that were prepared regularly interrogate an ISP’s server and use just the advertised off-hours capacity.

In closing

People liked fixed price deals. Unlimited is great, but there’s plenty of experience with bundles of minutes and the idea of data bundles has already showed up in 3G mobile data plans. The combination of several tiered data bundle prices with the availability of “free” connectivity for “on-net” peers and during off peak intervals is likely to appeal to file sharers and produce better results for both the sponsoring ISPs and file sharers alike.