OOKLA / Speedtest beta

eXDee: I've always had concern with how small the files truenet tests with. They tend to be 1 to 5MB. I see the test results hardly get close to showing the download potential of a 200mbps connection, even nationally. I can download from a university at the other end of the country and be close to saturating the line but I believe thats because TCP has time to ramp up. With truenet i can download off a nearby city and see only just over half of that 200mbps.

We have just 8 x 200Mb/s panelists, 5 of whom allow 5MB file downloads regularly, so while we do include some detail in the technical data we avoid commenting on ISPs for these. We download from both Wellington and Auckland and choose the best run from each server, so the comments below refer to the results from the best server.

Our 1MB download, while good for up to 100Mb/s, is not much use for a 200Mb/s connection so we avoid using it, we instead use the 2MB file, however we frequently review the results, because some probes still do not rampup in that time. However, 1 connection ramps up perfectly well inside 0.5MB. BTW for our 5MB file, 2 of 5 connections consistently rampup to 200MB on average without problems, one is yours. However 2 only make 150, rampup clearly completed in the 6th decile and one only makes 100Mb/s, rampup completed in the 5th decile.

We check all connections for speed changes every month, often asking the panelist. (the 100Mb/s one above is often checked)

Rampup is a topic I would love to find time to investigate further to comment on whether poor rampup is related to the ISP or if it is based on any other parameters - e.g. routes. The question could become: What is the correct speed, after rampup or including rampup? Or indeed a new test result; How fast is rampup?

We discovered a new problem today, with a 50MB connection registering 100MB/s on our tests, but on deeper inspection and with help of the panelist, we found that it is likely to be the shaping used to achieve 50Mb/s - the speed varies between 50 and 100+ Mb/s on a continuous basis, often 3-4 times in a 5MB file, averaging 50MB/s over a long download as the user does see on Speedtest. Our methodology is to pick the fastest quartile, so it always picks the 100Mb/s for this connection - clearly that is wrong according to the specs, but is it in practice? That becomes an issue as to how to interpret the information rather than a testing issue.

eXDee

JohnButt