[nitrogen]

In my area, Comcast runs its own Speedtest.net server to hide the fact that the closest peering point that leads back to the same state is a few thousand miles away and is insanely oversubscribed.

If Speedtest had more integrity, ISP-owned servers would be deprioritized.

[FireBeyond]

Not just that, Ookla /actively partners/ with the ISPs. Comcast's site is actually at speedtest.comcast.net.

There's a financial dis-incentive for them to do so.

[zokier]

Of course this is not all that dishonest. It is just a matter of what you want to measure. For last-mile speed measurement (which is what I personally use speedtest.net most for) using ISP provided local servers makes a lot of sense because it also removes most external factors from the measurement.

And really there is no "central point of Internet" which to measure speed to/from.

[nitrogen]

When the last mile was usually the limiting factor, last mile speed was the most important thing to measure. Now that major ISPs like Comcast are either throttling streaming video or routing streaming video through highly suboptimal connections, we need more speed tests that indicate real-world performance.

That said, I think my mention of integrity in my previous comment was excessive.

[FireBeyond]

How many servers on the Internet sit on the same switch on the same data center from your ISP's cable modem head end, though?