Concerned that broadband network customers in New York might not be getting the bandwidth they’re paying for, New York Attorney General Eric Schneiderman has decided to investigate.
His investigation apparently began as a result of some customers of broadband providers in the state who thought their Internet access was too slow, but he doesn’t say that in his press release asking broadband users to help with the investigation.
What Schneiderman does say is that he’s contacted three broadband providers in New York—Verizon, Cablevision and Time Warner—and has asked them for details on the actual broadband speeds delivered to customers.
In addition, the AG has arranged for dedicated access to internethealthtest.org (which is run by M-Lab), where customers can test Internet access throughput speeds themselves. The AG’s office then asks volunteers to fill out a form with the speeds they’re paying for, and to attach a screen shot with the results of the test.
However, despite the best intentions of the AG, this study does not provide the full picture of the various factors that affect broadband throughput. In fact, some of the fault (if there is any) that the AG is searching for may be beyond the control of the ISP.
I tried out the New York version of internethealhtest.org even though my office is located in Northern Virginia. While I was at it, I also ran the test without using the special New York settings, which basically means that the test uses the nearest server and not necessarily a server in New York. What the tests showed me was that my average throughput using Verizon FiOS is somewhat more than what I’m paying for, regardless of whether I’m accessing the server in Washington, D.C., or the one in New York.
But the test also revealed something that’s a lot more interesting. In each case, internethealthtest.org used a variety of third-party backbone providers to reach the destination server. Each time I tried, the choice of backbone providers changed, and in each case, one of the providers, usually Tata, was dramatically slower than the others. That in turn meant that my average throughput was slower than it might have been if the test had used broadband providers that had approximately equal throughput.
The average throughput speeds varied the most when one provider, again usually Tata, was pathetically slow, on the order of 12 megabits per second, even when the others were providing speeds on the order of 75M bps. However, it’s worth noting that Tata wasn’t always the slowest. If you run the test enough times, you’ll find that the backbone provider with really slow speeds can vary over time.
Adding to the complexity of this study, it turns out that the internethealthtest.org test doesn’t work on every browser, so the AG may well find that some of its volunteers can’t run the test at all.
New York AG’s Broadband Speed Test Results Won’t Tell Complete Story
When I tried it, the test ran properly on the Chrome and Firefox browsers, but not at all on Microsoft Edge, Internet Explorer or on Apple’s Safari for Windows. All tests were performed using a computer running Windows 10.
The series of tests I ran also revealed a fundamental flaw in the Attorney General’s understanding of how the Internet works. While broadband customers located in a major metropolitan area, such as Manhattan will likely have access to the test server using their provider’s network most of the way, that’s only likely if you are on the island of Manhattan.
If you’re elsewhere in New York (or like me, not in New York), then you’ll almost certainly have to depend on a third-party broadband provider for most of the trip between wherever you happen to be and where the server in New York City is located.
What’s more, while internethealthtest.org is able to make the details of the backbone provider transparent, this isn’t an option for Internet users under other circumstances. In addition, there’s really no way that an end-user can choose what backbone provider their traffic travels over. This means that even though one backbone may give great bandwidth, your traffic may not always use that one even if it’s available.
And that’s the problem with the AG’s test assumptions. The way TCP/IP traffic works is that when a packet gets to a router it chooses a path for a given packet according to what the router sees as an available connection at that specific moment. What this means is that the router may send your traffic over, say, Network X because the far-end router is advertising it as being open, even though Network Y is faster.
What’s more, in the real world, one stream of data may not follow the same path in its entirety. While it’s convenient to think of TCP/IP traffic as operating as if it’s switched, it isn’t. Your view of a Web page may have traveled over two or three different backbones only to be assembled when it gets to you. This, by the way, is a major cause of poor Internet experiences, but one that’s not really being addressed by the AG.
Of course, as M-Lab notes in their background information, interconnect arrangements between ISPs and backbone providers can also have an effect, as can the level of equipment that’s in place for any particular provider. Poor router choice and poor configuration can degrade throughput and that part is under the control of the provider.
The investigation by the New York AG may turn up some problems with promised throughput, and some of those problems may be due to the nature of the agreements between the three ISPs being investigated and their backbone providers.
I just hope the lawyers who are running this probe understand those basic realities of networking, or it could end up hurting Internet access in New York instead of helping.