Many network interconnection arrangements are deliberately kept confidential. On the other hand, some are not.
After the latest hubbub over whether Internet service providers are degrading traffic at certain network exchange points, industry analyst Dan Rayburn went straight to the company in question -- GTT Communications Inc. -- and asked why it was experiencing performance issues in some locations, but not others. As it turns out, GTT has not turned up all of its capacity with some ISPs, and it's turning down capacity in some places as it shifts network flow to accommodate recent infrastructure acquisitions.
In short, the performance issues are not the fault of the ISPs involved. Instead, GTT has made decisions internally that are affecting the results of tests taken through the Internet Health Test site. (See No Heroes in the Battle for the Net.)
The fact that there is no net neutrality issue with GTT doesn't change the issue of interconnection agreements in general. I still question how ISPs determine who should be allowed to connect to their last-mile networks and at what cost. However, given the politics involved (think Netflix Inc. (Nasdaq: NFLX) and the Federal Communications Commission (FCC) ), it would be virtually impossible for any outsider to determine what's fair and what isn't.
There was a brief moment of hope that the industry would find a way to standardize a unit of measurement in network interconnection deals. Last year Level 3 Communications Inc. (NYSE: LVLT) was working on a concept called bit-mile peering, which calculated traffic exchange costs based on both data volume and distance traveled. (See Net Neutrality: Level 3 Sees Peering Progress Soon.)
The company hasn't abandoned the concept, but it doesn't appear to have gotten much traction with ISPs either. Worse, now that Level 3 has signed new long-term interconnection deals with a number of last-mile providers, it's hard to see how there would be much incentive to keep pushing the bit-mile peering idea. (See Interconnect Deals Bear Net Neutrality's Stamp.)
In multiple conversations, Rayburn has maintained that the number of companies involved in interconnection deals with ISPs will always be small because almost all content providers run their content through a handful of content delivery networks (CDNs), and those CDNs have every incentive to make deals with ISPs to ensure high quality of service. In other words, even as video increases, most of it will still filter through a few select CDN providers, and as long as those companies and transit providers can forge agreements with ISPs, that's all that matters.
However, I wonder at how networks might evolve further, and whether we'll be looking at this same issue in a few years even while the latest round of long-term interconnection deals is still in place. What if localized content providers crop up that want to dump massive video traffic directly on to last-mile networks in a small geographic region? A provider like that might not need a CDN, but prefer instead to deal with ISPs directly.
As far as fairness goes in interconnection deals, Rayburn also points out that information disclosure goes both ways. Netflix and Level 3 for example, have both complained about the prices they've had to pay to ISPs, but neither has been willing to say how much money has actually exchanged hands. Without that information, it's impossible to judge whether costs are being shared equitably.
For the moment, Rayburn notes that any industry observer can access certain information that provides at least a little more context around the interconnect issue. Many network operators publish their peering policies (See Comcast Corp. (Nasdaq: CMCSA, CMCSK)'s policy here), and information on pricing for data transit is readily available.
Just like the results of the Internet Health Test, those data points aren't enough to paint a complete picture of how interconnection deals are done. But unless and until all network operators are willing to share details the way that GTT has, it's all we've got.
— Mari Silbey, Senior Editor, Cable/Video, Light Reading