Integrate Composite Times into Telemetry

RESOLVED FIXED in mozilla37

Status

()

Toolkit
Telemetry
RESOLVED FIXED
3 years ago
3 years ago

People

(Reporter: mchang, Assigned: mchang)

Tracking

unspecified
mozilla37
x86
Mac OS X
Points:
---

Firefox Tracking Flags

(Not tracked)

Details

Attachments

(1 attachment)

(Assignee)

Description

3 years ago
Track and measure composite times in telemetry.

Comment 1

3 years ago
Adding for reference bug 1048667 comment 132 and 133 (discussion on what metrics to measure).
(Assignee)

Updated

3 years ago
Blocks: 1107733
(Assignee)

Comment 2

3 years ago
Created attachment 8532341 [details] [diff] [review]
Integrate Composite Times into Telemetry

Integrate composite times into telemetry using a histogram. Set a cap at a 1 second maximum measurement to composite.
Attachment #8532341 - Flags: review?(bgirard)
This looks ok but wont be useful for Silk as its written.
Comment on attachment 8532341 [details] [diff] [review]
Integrate Composite Times into Telemetry

Looks good but please get a review on Histograms.json changes. I asked in #perf but I haven't gotten a response yet. I wonder if the high of 1000 is too high for composite times which should rarely be in that range if it's at the cost of precision.
Attachment #8532341 - Flags: review?(bgirard) → review+
(Assignee)

Comment 5

3 years ago
Comment on attachment 8532341 [details] [diff] [review]
Integrate Composite Times into Telemetry

From comment 4, asking :avih to review. He said the exponential scale should be good. Here is an example with a range of 0-10K and 50 buckets:

http://telemetry.mozilla.org/#filter=nightly%2F37%2FFX_REFRESH_DRIVER_FRAME_DELAY_MS&aggregates=multiselect-all!Submissions!Mean!5th%20percentile!25th%20percentile!median!75th%20percentile!95th%20percentile&evoOver=Builds&locked=true&sanitize=true&renderhistogram=Table
Attachment #8532341 - Flags: review?(avihpit)
(In reply to Benoit Girard (:BenWa) from comment #4)
> Looks good but please get a review on Histograms.json changes. I asked in
> #perf but I haven't gotten a response yet. I wonder if the high of 1000 is
> too high for composite times which should rarely be in that range if it's at
> the cost of precision.

Depends on what kind of granularity you need.. This is another exponential histogram with high of 1000  and 50 buckets: http://mzl.la/1sdX4RN
That's look good. It would better if 15,16,17 weren't shared at all but the resolution should be fine.

Updated

3 years ago
Attachment #8532341 - Flags: review?(avihpit) → review+
(Assignee)

Comment 8

3 years ago
https://hg.mozilla.org/integration/mozilla-inbound/rev/731e8e1a0494
https://hg.mozilla.org/mozilla-central/rev/731e8e1a0494
Status: ASSIGNED → RESOLVED
Last Resolved: 3 years ago
Resolution: --- → FIXED
Target Milestone: --- → mozilla37
You need to log in before you can comment on or make changes to this bug.