Add new memory histograms which fix inprecise bucketing.

Add new memory histograms which fix inprecise bucketing. Current Memory.Total and Memory.Swap.Total had a cap of 1000M which is very easy to exceed (especially due to shared memory overestimate).

BUG=516748
Committed: https://crrev.com/b8d7b276a310412e6bb1b12f6e46f1ddb1b24fa2
Cr-Commit-Position: refs/heads/master@{#341905}

[rkaplow, 2015-08-04 18:30:21]

rkaplow@chromium.org changed reviewers:
+ asvitkine@chromium.org, hajimehoshi@chromium.org, kouhei@chromium.org

[rkaplow, 2015-08-04 18:30:21]

As per an old internal discussion - will keep the old metrics around for a
milestone.

[Alexei Svitkine (slow), 2015-08-04 18:37:44]

https://codereview.chromium.org/1270193002/diff/40001/tools/metrics/histogram...
File tools/metrics/histograms/histograms.xml (right):

https://codereview.chromium.org/1270193002/diff/40001/tools/metrics/histogram...
tools/metrics/histograms/histograms.xml:17302: +<histogram name="Memory.Total2"
units="MB">
Should the unit be MiB?

[rkaplow, 2015-08-04 19:50:26]

On 2015/08/04 18:37:44, Alexei Svitkine wrote:
>
https://codereview.chromium.org/1270193002/diff/40001/tools/metrics/histogram...
> File tools/metrics/histograms/histograms.xml (right):
> 
>
https://codereview.chromium.org/1270193002/diff/40001/tools/metrics/histogram...
> tools/metrics/histograms/histograms.xml:17302: +<histogram
name="Memory.Total2"
> units="MB">
> Should the unit be MiB?

done

[Alexei Svitkine (slow), 2015-08-04 20:24:22]

lgtm, but please wait other reviewers' look over as well

[kouhei (in TOK), 2015-08-05 02:54:24]

Would you add more CL description? How was the old bucketing "inprecise"?
Probably:
- The cap of 1000MB wasn't enough
- We need more fine-grained bucket for better understanding.

I'm not sure if 100 linear bucket for 64000MiB is enough though. Can we increase
the number?

[rkaplow, 2015-08-05 03:16:17]

On 2015/08/05 02:54:24, kouhei wrote:
> Would you add more CL description? How was the old bucketing "inprecise"?
> Probably:
> - The cap of 1000MB wasn't enough
> - We need more fine-grained bucket for better understanding.
> 
> I'm not sure if 100 linear bucket for 64000MiB is enough though. Can we
increase
> the number?

The issue was mainly the 1000M cap and not the fine-grained bucketing. 
This actually uses the default which has exponential bucketing, and so even
though I've increased the max by a lot, becuase this is exponential and I've
increased #buckets from 50 to 100, this will actually end up giving more fine
grained buckets, even at the lower end.

[kouhei (in TOK), 2015-08-05 03:41:30]

On 2015/08/05 03:16:17, rkaplow wrote:
> On 2015/08/05 02:54:24, kouhei wrote:
> > Would you add more CL description? How was the old bucketing "inprecise"?
> > Probably:
> > - The cap of 1000MB wasn't enough
> > - We need more fine-grained bucket for better understanding.
> > 
> > I'm not sure if 100 linear bucket for 64000MiB is enough though. Can we
> increase
> > the number?
> 
> The issue was mainly the 1000M cap and not the fine-grained bucketing. 
> This actually uses the default which has exponential bucketing, and so even
> though I've increased the max by a lot, becuase this is exponential and I've
> increased #buckets from 50 to 100, this will actually end up giving more fine
> grained buckets, even at the lower end.

lgtm

Thanks! Sorry I didn't notice that this was exponential.

[hajimehoshi, 2015-08-05 04:10:28]

lgtm

[rkaplow, 2015-08-05 14:16:45]

The CQ bit was checked by rkaplow@chromium.org

[commit-bot: I haz the power, 2015-08-05 14:16:59]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1270193002/60001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1270193002/60001

[commit-bot: I haz the power, 2015-08-05 15:56:53]

Committed patchset #4 (id:60001)

[commit-bot: I haz the power, 2015-08-05 15:57:50]

Patchset 4 (id:??) landed as
https://crrev.com/b8d7b276a310412e6bb1b12f6e46f1ddb1b24fa2
Cr-Commit-Position: refs/heads/master@{#341905}