Interconnection: Towards a New Regime

[Commentary] Internet interconnection usually doesn’t make for big news. At an event hosted by the Congressional Internet Caucus Advisory Committee, David Clark, noted Internet engineer and MIT researcher, presented preliminary results from a joint MIT -- University of California at San Diego study on the causes and locations of congestion within the core of the Internet.

The researchers were cautious and reiterated that their results are preliminary, but the conclusions of their abstract read as follows:

• Our data does not reveal a widespread congestion problem among the US providers.
• Most congestion we see can be attributed to recognized business issues, such as interconnection disputes involving Netflix. These issues are being resolved, if slowly.
• Congestion does not always arise over time, but can come and go essentially overnight as a result of network reconfiguration and decisions by content providers as to how to route content.

There are some key take-aways from this new data.

First, it is worth noting the point made in Dr Clark’s presentation that there are numerous reasons a broadband user could have a frustrating experience with their broadband, and interconnection congestion is only one of them.

More importantly, congestion in the core of the network is rare and where it does occur it is because of real disparities between capacity and demand and not problems with the technology.

One arrangement that likely makes the most economic sense for a service that uses extreme amounts of bandwidth like Netflix would be to directly interconnect with last-mile networks. Direct interconnection would likely lower their costs and certainly improve their customer’s experience. Note that this is exactly what Netflix did.