Question

I have experimented with a technique that involves a web page request which you use to calculate the bit-rate based on bytes divided by the time elapsed. You can average multiple data points of course, but is this as accurate of a bit-rate estimation that can be made?

Do any professional or less hackish techniques exists? (Or is this just one of the magical mysteries of the internet?)

Answer 1

This really depends what kind of speed information you are interested in. You must be aware of several things:

1. The speed test may be skewed by caching
2. CDN servers that catch a request (assuming you are checking for HTTP response speed)
3. The offchance of the network traffic veering to a congested network
4. DNS caching

My suggestion is to make requests to a fair amount of unique sites for it to test. [Within an accetable margin of error]

Test lots of them and take the average speed, that is what you can generally expect from a client. The max is an estimated maximum that a client can get. The minimum that a client can guarentee is 0mbs.