Have you ever seen one of those roadside displays that say “Your speed is …?”
The displayed speed changes every time a car drives by because some drivers are fast while others drive more slowly. It is possible over time to determine an average speed of the traffic flow, but it is absolutely impossible to know the speed that all traffic is moving because not all the traffic is moving at the same speed. The information superhighway we call “The Internet” works exactly the same way, thus there is not, and cannot be, any accurate way to state definitively how fast an internet connection is all the time.
That last sentence is worth repeating because it is important that everyone understands.
There cannot be any accurate way to state definitively how fast an internet connection really is, as any connection speed constantly changes. Measurements of internet bandwidth speed are therefore averages or theoretical. They are not actual since the speed varies so much from second to second. In spite of this fact, consumers still want to know that they are getting all of the connection speed their Internet Service Provider (ISP) is supposed to be delivering to them. Hence they often turn to one of the many “speed test” web sites in an attempt to measure the speed of their connection. There they are shown test results they assume to be accurate. In reality their real data throughput speed may be something quite different. Anyone who takes time to run one of these speed tests a few dozen times over a period of hours or days and average the results will arrive at a reasonably accurate average. Relying, however, on the wrong test, a small sampling of tests, or a single test is inherently inexact.Most speed test sites report speed three ways: download speed, upload speed, and ping time. Download and upload speeds usually are reported in kilobits per second (Kbps) or megabits per second (Mbps). Ping times are measured in milliseconds (thousandths of a second). These statistics usually are computed based on the transfer of one small file over one connection from a webserver to your PC and can be accurate within that narrow context. What makes those figures inaccurate is assuming they apply to all situations.
Web browsers have download accelerators built in and use two or more connections. Thus the test using one connection is in no way an accurate measure of what a web browser does in normal usage. The speed test uses the TCP/IP protocol and does not test the UDP protocol used by VoIP and video software such as Skype.
Another big factor that skews speed tests is caching pages on local servers. If 100 gringos in Mexico read The New York Times it is very likely the local ISP may keep a copy of that web page on its local server so that the next time someone wants the Times it can be delivered to them not from the server in New York but from a nearby server in Mexico. Caching popular web pages in this way improves speed and minimizes duplicate internet traffic. Of course this also can affect the results of some popular speed test sites skewing the results. You might think you are checking the speed of a connection from your house in Mexico to Seattle. In reality that test might be reporting results only to Mexico City even when the test page says your speed was checked through to Seattle. The test might report a faster speed because it did not go all the way to Seattle.
It is important to understand that caching is great for many reasons, and Internet Service Providers do this to improve the speed and reliability of your online experience. Without caching your internet really would be slow. When it comes to speed tests, though, caching can make it difficult to know if your speed testing results are accurate or not.
Charles Miller is a freelance computer consultant with more than 20 years IT experience and a Texan with a lifetime love for Mexico. The opinions expressed are his own. He may be contacted through his web site at SMAguru.com.