Once upon a long ago, I used to swear by speedtest.net for broadband speed tests but then I discovered that their test method is not very scientific. Yesterday I aquired a new broadband connection from Etisalat and decided to do a four way speed comparision between Mobitel, Dialog, Lanka Bell and Etisalat, out of laziness I chose speedtest.net for the test, but the results from the Mobitel test against the server in Colombo reminded me that it's not such a good idea.
Spot it yet? It's not the slow 0.58mbps speed that I am worried about but the 3 second latency for the ping. There's no way that it can be so. The server is less than 10 kilometers from here and the server is hosted by Sri Lanka Telecom (SLTL) - the parent company of Mobitel. The test was repeated three times and on each occaision the results were similar.
Now let's see what a test using the Etisalat connection produces:
Notice how the Etisalat link appears to be faster than the Mobitel connection, well that's to be expected the Mobitel link is 1mbps while Etisalat is supposed to be 21mbps (but the USB modem in this case only supports speeds upto 3.6mbps) but notice how much faster the ping is.
Is Lanka Bell really all that slow? Nay not so. It's actually quite fast, even the Telecommunication Regulatory Commission concluded last year that their speeds are closest to the advertised speeds.
Thus it can be concluded once again that using speedtest.net is not a very scientific way of doing this.
