Here's a graph of some data about the differences between swarmed and non-swarmed bots: https://chromeperf.appspot.com/report?sid=54521a38b4bf18cafdf612dee67ed05223f588049cce73d2cad50a38a7e9c5ee

Things look mostly the same.

+v8 experts: we are working on migrating Android devices in our perf lab to swarming ( issue 670879 ). Part of the process requires us to validate that the new infrastructure won't make our perf data noisier/more overhead. 

Based on Stephen's graph in #2, looks like Android swarming is less noisy than the current bot? Can any of you helps validating? Also besides speedometer, does v8 team want us to compare the data of other benchmarks?

Assign to ulan@ for further triaging.

The graph in #2 looks great to me.

Are you going to check the system-health benchmarks too? If noise does not increase there, then I think we should be good from V8 point of view.

Yes, we also gonna check the system_health benchmarks. In specific:
system_health.memory_mobile. We don't check system_health.common_mobile because the metrics there are still pretty much in flight (the only productionized one is power & we don't have battor on these swarming bots yet, AFAIK).

Looks less flaky on RCS benchmarks too: https://chromeperf.appspot.com/report?sid=1eb9685868adb97aeda928fc7d9ef771819d02f75b990974f756d0ac3602e38e

so, +1 what ulan@ said.

Ok, I'm marking this as Fixed, since you say this metric looks good to you.