Closed Bug 1045071 Opened 10 years ago Closed 9 years ago

Power statistics

Categories

(Testing :: General, defect)

x86_64
Windows 7
defect
Not set
normal

Tracking

(Not tracked)

RESOLVED WONTFIX

People

(Reporter: ekyle, Unassigned)

References

Details

This bug is about tracking any changes that might be needed to deal with the ammeter measurements.  Here is the email that initiated this:

On 17/07/2014 9:56 AM, Jonathan Hylands wrote:
>
> Hi Kyle,
>
> I don't remember if we've ever met, but I've been working on power tools
> for Firefox OS over the past year. I'm in the process of setting up a
> performance server that will run power tests, and we want to push the data
> to Datazilla.
>
> A typical test will put the phone in a specific state, and then measure
> power usage over a 30 second period. During that time, we will gather
> between 3500 and 3600 current (mA) samples from the ammeter. I average them
> together to get a single value, but when we push the data to Datazilla we
> want to include more than a single sample. Typically when we use b2gperf
> for timing, we end up with 30 samples per data point.
>
> So, the question is, is 3600 samples too many? Jeads told me that the data
> structure can handle 5000, so from that perspective it should be okay, but
> I'm not sure whether the GUI will handle it nicely, or if it will slow
> things down too much.
>
> Is there a reasonable statistically-valid way to turn 3600 entries into a
> much smaller number, like 30 or 60?
>
> Thanks,
> Jon
Since it appears the ammeter is being used to measure steady state, not events, and both the voltage and duration are constant, mean is a fine statistic to use.
Blocks: 962378
No longer blocks: 1014275
Component: Bugzilla Anthropology Metrics → General
Product: bugzilla.mozilla.org → Testing
Version: Production → unspecified
On 16/09/2014 3:02 PM, Joel Maher wrote:
> one way to get all the revisions would be to query all the revisions from:
> https://datazilla.mozilla.org//refdata/pushlog/list/?days_ago=7&branches=Firefox (I assume this is done for Talos)
>
> Iterating through the list, there is an id and id['revisions'] yields a list of revisions in that specific push.  Pulling the last item from the list <id>['revisions'][-1], yields the revision to use:
>
>  * https://datazilla.mozilla.org/power/testdata/raw/mozilla-central/cd918760da39?product=Firefox&os_name=Win&os_version=7%20-%20Firefox&branch_version=35.0a1&processor=x86&build_type=opt&test_name=PowerGadget <- returns [] as we didn't run tests on this
>
>  * https://datazilla.mozilla.org/power/testdata/raw/mozilla-central/bf5fcc0c4b27?product=Firefox&os_name=Win&os_version=7%20-%20Firefox&branch_version=35.0a1&processor=x86&build_type=opt&test_name=PowerGadget <- returns a full set of data.
>  ** note, to get IE data, change the os_name as in this example: https://datazilla.mozilla.org/power/testdata/raw/mozilla-central/bf5fcc0c4b27?product=Firefox&os_name=Win&os_version=7%20-%20Internet%20Explorer&branch_version=35.0a1&processor=x86&build_type=opt&test_name=PowerGadget 
>
> Maybe there is a better way to go about this, maybe not!
Kyle, just to be clear, Jon's power suite and mine are separate things. He is measuring power usage for Firefox OS while I measure it for Firefox and we have different tools with different formats.
dzAlerts is a dead project
Status: NEW → RESOLVED
Closed: 9 years ago
Resolution: --- → WONTFIX
You need to log in before you can comment on or make changes to this bug.