Add UMA histograms for the inter-process TimeTicks skew.

Add UMA histograms for the inter-process TimeTicks skew.

This will give us visibility into adjustments made through
InterProcessTimeTicksConverter when putting together records for
Navigation Timing API.

BUG=394757
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=290241

[ppi, 2014-08-15 16:39:51]

Hello

Alexei, Charlie - ptal.

As described in the crbug, we don't adjust |navigationStart| TimeTicks stamp
when passing it around between processes. This should tell us how worried we
should be about it.

Also, it will allow to validate if the conversion works as expected and sompare
TimeTicks skew across platforms.

[Alexei Svitkine (slow), 2014-08-15 16:54:34]

lgtm % comments

https://codereview.chromium.org/481443003/diff/1/content/child/resource_dispa...
File content/child/resource_dispatcher.cc (right):

https://codereview.chromium.org/481443003/diff/1/content/child/resource_dispa...
content/child/resource_dispatcher.cc:792:
"InterProcessTimeTicks.IsSkewAdditive_BrowserToRenderer", true);
Nit: Extract the if condition outside the if into a local var and log this
histogram only once outside the loop. (Will reduce number of macro invocations,
since each one expands to a bunch of code and adds to the binary bloat.)

Same for the other code.

https://codereview.chromium.org/481443003/diff/1/tools/metrics/histograms/his...
File tools/metrics/histograms/histograms.xml (right):

https://codereview.chromium.org/481443003/diff/1/tools/metrics/histograms/his...
tools/metrics/histograms/histograms.xml:10201: +<histogram
name="InterProcessTimeTicks.IsSkewAdditive">
Nit: Specify an enum="Boolean" or a more specific variant.

[Charlie Reis, 2014-08-15 17:05:24]

LGTM

[ppi, 2014-08-18 09:29:05]

Thanks!

https://codereview.chromium.org/481443003/diff/1/content/child/resource_dispa...
File content/child/resource_dispatcher.cc (right):

https://codereview.chromium.org/481443003/diff/1/content/child/resource_dispa...
content/child/resource_dispatcher.cc:792:
"InterProcessTimeTicks.IsSkewAdditive_BrowserToRenderer", true);
On 2014/08/15 16:54:33, Alexei Svitkine wrote:
> Nit: Extract the if condition outside the if into a local var and log this
> histogram only once outside the loop. (Will reduce number of macro
invocations,
> since each one expands to a bunch of code and adds to the binary bloat.)
> 
> Same for the other code.

Done.

https://codereview.chromium.org/481443003/diff/1/tools/metrics/histograms/his...
File tools/metrics/histograms/histograms.xml (right):

https://codereview.chromium.org/481443003/diff/1/tools/metrics/histograms/his...
tools/metrics/histograms/histograms.xml:10201: +<histogram
name="InterProcessTimeTicks.IsSkewAdditive">
On 2014/08/15 16:54:34, Alexei Svitkine wrote:
> Nit: Specify an enum="Boolean" or a more specific variant.

Done.

[ppi, 2014-08-18 09:29:12]

The CQ bit was checked by ppi@chromium.org

[commit-bot: I haz the power, 2014-08-18 09:29:49]

CQ is trying da patch. Follow status at
 https://chromium-status.appspot.com/cq/ppi@chromium.org/481443003/20001

[pasko, 2014-08-18 09:30:34]

Fly by review.

The idea is nice, however, I am interested in more details on what we can do
with somewhat unreliable data we'd get.

https://codereview.chromium.org/481443003/diff/1/content/common/inter_process...
File content/common/inter_process_time_ticks_converter.cc (right):

https://codereview.chromium.org/481443003/diff/1/content/common/inter_process...
content/common/inter_process_time_ticks_converter.cc:76: return numerator_ == 1
&& denominator_ == 1;
I guess non-additive cases are assumed to be low, is that right? What can we do
with the data once we get it?

After a quick first glance, the fitting in the Converter does not look very
scientific to me, which makes me wonder if it would skew our skew :) Are
Converters readjusted at some point in time? How many measurements are made to
get *_lower_bound and *_upper_bound before we conclude whether we are additive?

https://codereview.chromium.org/481443003/diff/1/tools/metrics/histograms/his...
File tools/metrics/histograms/histograms.xml (right):

https://codereview.chromium.org/481443003/diff/1/tools/metrics/histograms/his...
tools/metrics/histograms/histograms.xml:10188: +    Estimated additive skew
between processes, recorded if the browser process
It is usually a good idea to mention _when_ the histogram is collected.

[ppi, 2014-08-18 10:05:16]

Thanks for the remarks, Egor!

The data on the inter-process skew will be subject to skews coming from variable
IPC overhead, but we will get precise data on the adjustments we make to
compensate the skew, that is, on how the converter works in the field.

The converters are short-lived - every time we record UMA, it is one converter
that just has been created and will be deleted soon. That is, we don't take
subsequent measurements to calibrate it better. Which might be good given that
the inter-processor skew was reportedly changing in time on Windows.

[commit-bot: I haz the power, 2014-08-18 12:40:53]

Committed patchset #2 (20001) as 290241