Battery impact UMA

tools/metrics/histograms/histograms.xml

Index: tools/metrics/histograms/histograms.xml
diff --git a/tools/metrics/histograms/histograms.xml b/tools/metrics/histograms/histograms.xml
index 1ffa31496a7c71ef8fa5ab8aa3024bcddd138e9f..23729eb4c6b9dc9aa24d950e41889d3690f129bf 100644
--- a/tools/metrics/histograms/histograms.xml
+++ b/tools/metrics/histograms/histograms.xml
@@ -25187,6 +25187,16 @@ Therefore, the affected-histogram name has to have at least one dot in it.
   </summary>
 </histogram>
 
+<histogram name="Power.BatteryDischargePercentPerHour" units="%">
+  <owner>jeremy@chromium.org</owner>
+  <summary>
+    The amount of power used per hour. Reported every time the system is on

[Mark P, 2014/11/10 18:08:36]

Please make this description clearer.  The comment says "amount of power".  The
units say "%".  These are different things.  Furthermore, if you meant to be
talking about percent, what is the percent relative to? Starting battery power
level?  Full battery power level?

+    battery power for more than 30 minutes. This measurement is tied tightly to

[Mark P, 2014/11/10 18:08:36]

If the system is on battery power for more than 30 minutes yet sleeping, is this
recorded?

Also, is this reported every 30 minutes, or only the first time the system is on
battery power for more than thirty minutes?  i.e., if I keep using my laptop
after this this or another hour, do we report additional samples?

Furthermore, if the latter, then how do you handle the second report (the one
after being on battery for an hour)?  It is the last thirty minutes normalized
to be a "percent per hour" rate, or it is the actual percent per hour rate for
the last hour?

In short, these descriptions should be detailed.  You need to import part of the
comments you've answered in the code review thread and even in the changelist
description into these descriptions.

Many of these comment apply to below histograms too.

+    hardware model/OS and is not camparable across differently hardware
+    configurations.
+  </summary>
+</histogram>
+
 <histogram name="Power.BatteryDischargeRate" units="mW">
   <owner>derat@chromium.org</owner>
   <summary>
@@ -25195,6 +25205,33 @@ Therefore, the affected-histogram name has to have at least one dot in it.
   </summary>
 </histogram>
 
+<histogram name="Power.BatteryDischargeRate_15" units="%">
+  <owner>Jeremy@chromium.org</owner>

[Mark P, 2014/11/10 18:08:36]

nit here and below: please don't captizalize usernames

+  <summary>
+    The percent of depleted battery charge over the first 15 minutes after the
+    system goes on battery power. This measurement is tied tightly to hardware
+    model/OS and is not camparable across differently hardware configurations.
+  </summary>
+</histogram>
+
+<histogram name="Power.BatteryDischargeRate_30" units="%">
+  <owner>Jeremy@chromium.org</owner>
+  <summary>
+    The percent of depleted battery charge over the first 30 minutes after the
+    system goes on battery power. This measurement is tied tightly to hardware
+    model/OS and is not camparable across differently hardware configurations.
+  </summary>
+</histogram>
+
+<histogram name="Power.BatteryDischargeRate_5" units="%">
+  <owner>Jeremy@chromium.org</owner>
+  <summary>
+    The percent of depleted battery charge over the first 5 minutes after the
+    system goes on battery power. This measurement is tied tightly to hardware
+    model/OS and is not camparable across differently hardware configurations.
+  </summary>
+</histogram>
+
 <histogram name="Power.BatteryDischargeRateWhileSuspended" units="mW">
   <owner>derat@chromium.org</owner>
   <summary>
@@ -25206,6 +25243,15 @@ Therefore, the affected-histogram name has to have at least one dot in it.
   </summary>
 </histogram>
 
+<histogram name="Power.BatteryDischargeReportsCancelled" units="count">
+  <owner>jeremy@chromium.org</owner>
+  <summary>
+    The number of times an error was detected reporting Power.BatteryDischarge_*
+    due to the actual delay being much greater than the expected delay in taking
+    the measurement.

[Mark P, 2014/11/10 18:08:36]

This histogram doesn't seeming meaningful without a comparison point.  If I see
the number "500" here, how do I decide if that's a lot?  Which histogram do I
compare it to?  Suppose the other histograms have counts (respectively of 5000,
2500, 2000, 1500)?  I would think you need to break errors down by which
histogram we were attempting to record and then failed at recording.

Also, is the "delay" the time it takes to evaluate the power usage or the
duration that the histogram is supposed to be recording over?  Please clarify.

If it's the later, is "much greater" relative to the size of the expected
duration?  It is, say, a deviation by 10% or an absolute deviation like say 10
minutes.

+  </summary>
+</histogram>
+
 <histogram name="Power.BatteryInfoSample" enum="BatteryInfoSampleResult">
   <owner>derat@chromium.org</owner>
   <summary>