I want to do a general review of mobile site speed, so it doesn’t really matter if I’m testing for iOS, Android, etc. I did one test from Dulles and it was absurdly slow (slightly over 24 seconds for full load).
I would like to test from a few different locations in the US, but noticed that Dulles seems to be the only location with which I can test a mobile browser.
My questions are:
- Is using a mobile browser the best way to test mobile latency?
- Do I need to use the Advanced Settings and select a mobile 3G connection?
- What does dynaTrace mean, and is it relevant to my situation?
The mobile devices in Dulles are all behind a fixed 3G connection and it’s not configurable. You can try a few different locations on Akamai’s mobitest: http://mobitest.akamai.com/m/index.cgi but I believe most of those are connected wifi to a fast internet connection.
Do you have a link to a test result? 24 seconds may be accurate depending on how the page is constructed.
You can also use Chrome from any location with a few settings tweaks to get it to load your mobile test over a shaped 3G connection:
It’s not perfect but it will eliminate the actual device from the equation and give you the same detail you are used to getting from desktop testing.
Oh, and no, dynatrace is not relevant to your testing. Dynatrace provides profiling of javascript execution in IE (among other things).
This is really helpful. Thanks very much!
Have you considered RUM for sampling the load times of your mobile visitors?
No, I haven’t…but I’m not sure what RUM is. As you might imagine, a search turns up mostly mixed-drink-related results 
Sorry, RUM is Real User Measurement. It is basically the process of using javascript in the page to report the actual load times back to a server. Google Analytics also provides the information as part of their reporting if you use it (and there are other free and commercial services that do as well).
Good timing, I just wrote up this article which covers quite a bit of it: http://queue.acm.org/detail.cfm?id=2446236
iPad is 1024x768 so it depends if you want to test portrait or landscape. That’s the default resolution for desktop testing as well so if you’re testing landscape you don’t really need to change the viewport.
The retina iPads have a much higher pixel resolution but css scaling is 2 so it’s effectively the same resolution (unless you are trying to test retina images in which case you need a seriously high-resolution desktop to emulate it and you won’t be able to do it on any of the WebPagetest machines).