Internet providers caught inflating speed test results


So what about mobile networks? One on-going problem I have with the Irish Three mobile network is that its data performance gets ridiculously slow after about 5pm even on 4G. So I thought – Is it throttling or prioritising any ports? Whilst the variation was mixed in the evening, port 8080 test results are generally higher. The difference became very clear early in the day when there is less network traffic. So just after I arrived at my workplace, I ran two speed tests near a window with a full LTE signal, again HTTP vs port 8080:


Three speed test over http
Three speed test over port 8080

Unlike the fixed broadband providers, Three also appears to prioritise HTTPS traffic, delivering similar results to the port 8080 based tests, probably to improve video streaming performance. Three also delivers considerably higher test results with multithread tests which is also a fairly clear sign of high contention on their network. The following shows an example of HTTP vs HTTPS single thread tests conducted on the Three network one morning, one right after the other with the same 25MB block size:

Three speed test #2 over http
Three speed test #2 over https


I recently bought a prepay Vodafone phone SIM to see how Vodafone compares with the Three network, and decided to try some further tests with their 4G (LTE) network. They have very limited monthly download allowances included with their mobile data plans and the largest mobile broadband package they offer online is 30GB, compared to 250GB with Three. So I did not expect Vodafone to perform any traffic shaping on their network. However, the speed tests on TestMy with HTTP vs 8080 reveal an interesting surprise:

Vodafone speed test over http
Vodafone sped test over port 8080

Like Three, Vodafone also delivers much faster throughput on the HTTPS port also. However, unlike every other ISP I tested, I sometimes could not reach the test results was giving even when I ran multithread tests. So I looked at the TiP (Test in Progress) graphs from a few of the TestMy results over HTTPS and port 8080 and many of them had a graph similar to the following with several throughput dips intermittently spaced through the test:

Vodafone speed test Test in Progress graph

As mentioned in the test flow documentation, it discards the fastest 10% and slowest 30% of the 20 slices each thread is broken into, so with this single connection, it probably would have reported about 55Mbps based on the throughput dips accounting for roughly 30% of the test duration. As TestMy measured this 50MB transfer just like a stopwatch timing, its 42.1Mbps result is the average speed measured during this test. So not only was Vodafone delivering faster throughput over port 8080, it appears to have done it in such a way as to reduce data usage on its network with a minimal effect on the test result.

Due to the lack of 4G coverage on the Meteor network in my area, I have not been able to conduct any comparison tests on its 4G network. However, based on running speed tests over 3G, Meteor appears to behave properly, even at my workplace where its 3G network delivers up to 25Mbps down in the morning over the http port.

Let’s head to the next page where we take a look at what happens with Ookla’s Speedtest App when port 8080 is blocked…