Closed Bug 1811065 Opened 2 years ago Closed 2 years ago

[Experiment] Pref-Flip Experiment: Impact of Parallel Marking V2 Fx 111.0 to 112.0 Nightly

Categories

(Shield :: Shield Study, task, P3)

Product:
Shield
Component:
Shield Study
Type:
task

Tracking

(Not tracked)

Status:
RESOLVED FIXED

People

(Reporter: experimenter, Assigned: shell)

References

(Blocks 1 open bug,
URL
)

Details

User Story

Delivery Type: Pref Flip Experiment

    What is the preference we will be changing

javascript.options.mem.gc_parallel_marking

    What are the branches of the experiment and what values should
    each branch be set to?

- Control Default 50%:

Value: false

Default
        
- Treatment Parallel Marking 50%:

Value: true

Parallel Marking Enabled
        

    What version and channel do you intend to ship to?

100% of Nightly Firefox 111.0 to 112.0

    Are there specific criteria for participants?

Prefs:

Experiments:
Countries: all

Locales: all

    What is your intended go live date and how long will the experiment run?

Jan 18, 2023 - Feb 08, 2023 (21 days)

    What is the main effect you are looking for and what data will you use to
    make these decisions?

We will measure the impact on page load by measuring a change in the following GC and page load telemetry probes:

'perf_page_load_time_ms', 'time_to_first_interaction_ms',
'input_event_response_ms', 'input_event_response_ms_parent',
'perf_first_contentful_paint_ms',
'js_pageload_execution_ms', 'js_pageload_delazification_ms',
'js_pageload_parse_ms', 'js_pageload_protect_ms',
'js_pageload_xdr_encoding_ms', 'js_pageload_baseline_compile_ms',
'gpu_keypress_present_latency',
'fx_new_window_ms', 'fx_tab_switch_composite_e10s_ms',
'memory_total', 'memory_unique_content_startup',
'cycle_collector_max_pause', 'cycle_collector_max_pause_content',
'gc_max_pause_ms_2', 'gc_max_pause_ms_2_content',
'gc_ms', 'gc_ms_content',
'gc_slice_during_idle', 'gc_slice_during_idle_content',
'gc_mark_ms', 'gc_mark_rate_2',
'main_crashes_per_hour', 'content_crashes_per_hour',
'oom_crashes_per_hour', 'shutdown_hangs_per_hour'

    Who is the owner of the data analysis for this experiment?

None

    Will this experiment require uplift?

False

    QA Status of your code:

Not Started, QA request sent in https://mozilla-hub.atlassian.net/browse/QA-1712
Signed off as GREEN, a copy can be found here: https://docs.google.com/document/d/1Gvp8cr_nHVkK7BGFmMhuwfhcEbCFJagG9B_FZtAGzko/edit

    Link to more information about this experiment:

https://experimenter.services.mozilla.com/experiments/impact-of-parallel-marking-v2/

Impact of Parallel Marking V2

This experiment aims to assess the impact of parallel marking for the GC.

Experimenter is the source of truth for details and delivery. Changes to Bugzilla are not reflected in Experimenter and will not change delivery configuration.

Data Science Issue: https://mozilla-hub.atlassian.net/browse/DS-2508
More information: https://experimenter.services.mozilla.com/experiments/impact-of-parallel-marking-v2/

User Story: (updated)
Summary: [Experiment]: Pref-Flip Experiment: Impact of Parallel Marking V2 → [Experiment] Pref-Flip Experiment: Impact of Parallel Marking V2 Fx 111.0 to 112.0 Nightly
Start Date: 2023-01-18 End Date: 2023-02-08
Start Date: 2023-01-18 End Date: 2023-02-08
Status: NEW → RESOLVED
Closed: 2 years ago
Resolution: --- → FIXED

Looking at the results for the content process on machines with at least two CPU cores we see mark rate improvement of ~20%, leading to a decrease in marking time of ~15% and of total GC time by ~12%. Results are better on MacOS and Linux, with linux showing a decrease in marking time of 19% and total GC time of 15%.

User Story: (updated)
You need to log in before you can comment on or make changes to this bug.