Almost any time you access the Internet, your computer transmits and receives data that travels across not only your own Internet provider’s network, but also the networks of Internet backbone providers and other ISPs before reaching its final destination. This is what puts the “inter” in the “Internet”: independent networks connected together physically at interconnection points. But this same feature that makes the Internet so powerful can also be a weakness. If the interconnection between two networks doesn’t have the capacity to handle the traffic being sent across it, then that interconnection can become a point of congestion, leading to slower speeds and dropped packets.
A report released earlier this week published by M-Lab, a research consortium focused on measuring global Internet performance, studied this problem. They were curious how interconnection affected US Internet users’ connection quality, and how particular interconnections between “access” ISPs (e.g. Comcast, Verizon, Cox, Time-Warner Cable, etc.) and “transit” ISPs (Internet providers that connect between access ISPs and websites or other Internet services) were performing. Sadly, the results were not encouraging.
M-Lab’s data showed that the interconnections between certain access ISPs and transit ISPs were subject to consistent congestion, causing customers to experience sub-broadband speeds, more latency, and more dropped packets. In particular, for a period of nine months from May 2013 to February 2014, customers of Time-Warner Cable, Comcast, Verizon, AT&T, and CenturyLink all experienced degraded connections to the transit ISP Cogent.
For those following the net neutrality debate, this should be no surprise. It was during this period that transit ISPs like Cogent and Level 3 accused access ISPs like Verizon and Comcast of refusing to upgrade their interconnections to handle the increased traffic load, unless transit ISPs agreed to pay substantial fees above and beyond the cost of upgrading the interconnection infrastructure. In return, the access ISPs claimed that transit ISPs were at fault for trying to send so much traffic (primarily Netflix video) onto their networks.
While much of the congestion subsided once Netflix acceded to the access ISPs’ demands and paid up, the fact remains that for nine months, ISPs did not deliver the speeds they promised to many of their customers. As we explained in our reply comments to the FCC:
[B]y advertising a given speed to a customer, an ISP is essentially promising that customer that they will be able to move data at that speed through the ISP’s own network and to any network with which the ISP is peered. If the data transfer is slower because networks beyond the ISP’s are congested, that is not the ISP’s fault. But if the data transfer is slower than advertised because the ISP refuses to augment congested ports to a peer network, even when the peer network is willing to do so, then that congestion is solely the fault of the ISP.
But instead of upgrading their interconnection infrastructure (which one transit ISP even offered to pay for) access ISPs instead treated their customers like hostages in their negotiations with transit Internet providers.
In an ideal world, there would be enough competition among access ISPs that this situation would never have happened: if your traffic was experiencing congestion because your current Internet provider was refusing to interconnect with a certain provider, then you could simply switch to another ISP that was interconnected.
But we don’t live in an ideal world; we live in a reality where most large ISPs act as quasi-monopolies and changing ISPs comes with high switching costs. And of course, reality is also more complicated. After all, nobody expects an ISP to interconnect with every single organization that demands an interconnection. And nobody expects an ISP to enter into a commercially unreasonable peering contract.
Unfortunately, the one thing the M-Lab study can’t tell us is whether or not the contracts the transit ISPs suggested were unreasonable or whether the demands of the access ISPs truly were tantamount to extortion. This underscores the need for more transparency when it comes to peering and interconnection agreements. As we told the FCC, we believe such agreements should be made public so that transit providers and ISPs can see for themselves if they’re getting a fair deal, and so that customers can see if their Internet provider is causing congestion by making outrageous demands.
Without greater transparency, Internet customers will be in the same position the next time this happens: caught in the crossfire in the battle between access and transit ISPs.