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 1727ca2af1de172d509f753c4c577ae0f75a421b..e6686489cd50ba9a0741f42fe6347c663bd5968f 100644
--- a/tools/metrics/histograms/histograms.xml
+++ b/tools/metrics/histograms/histograms.xml
@@ -41900,6 +41900,13 @@ http://cs/file:chrome/histograms.xml - but prefer this file for new entries.
   </summary>
 </histogram>
 
+<histogram name="Platform.MeminfoMemTotal">

[Luigi Semenzato, 2016/09/19 17:03:09]

Instead of giving the units in the summary section, use the units specification
(see other histograms).

Also this may be a good time to reconsider the choice to re-send this number
every 30s instead of just once.  (Or file a bug for it.)

Thanks!

[jwd, 2016/09/19 17:28:02]

I was going to question the 30s idea. How are these metrics analysed?

What is "installed memory"? Is it the physical ram amount? If it's something
that doesn't change within a chrome session, you could be better served by
logging this once per report, you'd still be able to correlate it with other
metrics in each report, but you'd much reduce the counts magnitude. You'd also
be able to potentially normalize by UMA.RecordsCounts. (could introduce a bias
if CrOS tends to have varying intervals for each UMA log record, which the every
30s somewhat mitigates)

+  <owner>bccheng@chromium.org</owner>
+  <summary>
+    Chrome OS total installed memory in Kbytes (snapshotted every 30s).
+  </summary>
+</histogram>
+
 <histogram name="Platform.MeminfoMemUsedDerived">
   <owner>bccheng@chromium.org</owner>
   <summary>