Google maps 300TB of real-world Internet speed data

How fast is your broadband? M-Lab, a partnership between the New America Foundation and Google meant to measure Internet connections, has given Google two years worth of actual broadband connection data, as measured by users. That's more than 300TB of data, which Google has imported into its Public Data Explorer for easy viewing and analysis. The results are remarkable.

Measuring Internet access has been tricky for years. Sascha Meinrath of the New America Foundation told Ars back in 2009, when M-Lab got underway, that detailed network data about speeds, latency, jitter, and more used to be in the public domain until the government-run NSFnet was privatized in the earlier 1990s. Today, though, it's hard to know what speeds ISPs are actually offering (knowing what speeds they advertise, by contrast, is simple).

M-Lab has distributed testing tools for two years now and its servers have recorded data on the results. One of the most basic measurements is pure speed, measured in megabits per second. When these real-world speeds are charted on a map, they make Internet speed differences obvious in a way often obscured by simple lists and numbers. For instance, the two images below compare Internet download speeds in US states to Internet download speeds in European countries (many of which are the same size as US states). Speeds are medians.

US download speeds by state (red is highest, then orange, yellow, green, blue)

European download speeds

The data comes from the Network Diagnostic Tool (NDT), which Google says was "developed by Internet2 and widely deployed. The platform, the tool, and the data are all open—which means the Internet community can vet the measurement methodology, perform independent analysis of the same data, and build their own visualizations."

Zooming in on England's download speeds

The data can be drilled in on, going all the way down to the city level. In the UK, the government has already commissioned real-world speed testing and has the results; the same testing company, SamKnows, is currently running similar tests in the US on behalf of the Federal Communications Commission, though its results are not yet public.

M-Lab's results, while interesting, aren't based on random statistical sampling and so should be used with caution before drawing any policy implications. Still, at least among Internet users most likely to run a tool like NDT, European broadband doesn't look too shabby, especially in the north and center of the continent.