In There’s always something new to measure, we proposed several possible tests for the next version of WebXPRT. Of those, battery life testing generated the most interest.
Battery life testing poses a number of challenges. It’s not as simple as making WebXPRT loop. The biggest challenge is that different devices take different measures when the battery runs low. These measures range from dimming the screen, to stopping the hard disk, to totally shutting down the device. While these are perfectly reasonable, they are out of the benchmark’s control. Worse, most current browsers offer no way of knowing that these measures even happen nor do they offer good ways of querying the device to find out the state of its battery. We want to make sure that our approach does not unfairly advantage one device over another and gives a fair and accurate measure.
Because WebXPRT is a hosted application, we are looking at one of the other XPRT benchmarks for our first attempt at adding battery life to an existing benchmark. MobileXPRT seems to be the best fit. It runs on Android, which has a functional API for monitoring and managing power events, and the diversity of the Android ecosystem forces the benchmark to deal with a greater range of devices and OS configurations than TouchXPRT.
We are trying a number of approaches, and we have made some progress. We will discuss what we have learned in the next few weeks.
Our hope is that what we learn from MobileXPRT will better equip us to add battery life testing to WebXPRT.
Have any thoughts or comments? Post to the forums or e-mail benchmarkxprtsupport@principledtechnologies.com to let us know.
-Bill
