Ready for Trial: Quick intensive timer throttling of loaded background pages

[Zhang, Jiahe]

Contact emails

jiahe...@intel.com, fdo...@chromium.org

Specification

https://html.spec.whatwg.org/multipage/timers-and-user-prompts.html

Design docs

https://docs.google.com/document/d/1WFyfKUUxqM7uKxKOGhLiOjyY6T7QRduVcuHN0f6vJkk/edit

Summary

Enter Intensive Wake Up throttling after 10 seconds if the page is fully loaded when it becomes hidden. Currently, wake ups from JS timers with a nesting level >= 5 are throttled to 1 per minute after the page has spent 5 minutes in the background [1], which is very conservative and was chosen to allow a launch of Intensive Wake Up Throttling with minimal regression risk. We're now planning to reduce this timeout to 10 seconds if the page is fully loaded when hidden. [1] https://chromestatus.com/feature/4718288976216064

Blink component

Blink>Scheduling

TAG review

Not applicable. This feature changes the behavior of an existing API, while remaining spec-compliant ("Optionally, wait a further implementation-defined length of time.")

 

Risks

Interoperability and Compatibility

Gecko: No signal

WebKit: No signal

Web developers: No signals

Other signals: The more conservative version of Intensive Wake Up Throttling shipped smoothly to 100% Stable more than 1 year ago. A few bugs were filed, but in all cases we've been able to propose workarounds which made apps more efficient (example).

WebView application risks

Does this intent deprecate or change behavior of existing APIs, such that it has potentially high risk for Android WebView-based applications?

 

No, this feature will only ship on desktop platforms.

 

Goals for experimentation

We plan to experiment on 1% Stable to confirm whether we observe the same memory and power improvements as in the lab and on lower channels. We will decide whether this intervention ships based on the experiment data.

Ongoing technical constraints

None

Debuggability

This is not a new Web Platform feature.

Will this feature be supported on all six Blink platforms (Windows, Mac, Linux, Chrome OS, Android, and Android WebView)?

No

This feature will only ship on desktop platforms. On Android, the system severely limits resource consumption from background renderers, which makes this feature unnecessary.

Is this feature fully tested by web-platform-tests?

Yes

Flag name

quick-intensive-throttling-after-loading

Requires code in //chrome?

False

Tracking bug

https://bugs.chromium.org/p/chromium/issues/detail?id=1324656

Estimated milestones

DevTrial on desktop

105

Link to entry on the Chrome Platform Status

https://chromestatus.com/feature/5580139453743104

This intent message was generated by Chrome Platform Status.

 

 

[François Doray]

+Scott Haseley as an expert in this field.

We would like to start experimenting with this intervention on 1% Stable very soon. We've been experimenting on 50% of Beta for almost 2 months. The results are encouraging and we aren't aware of negative Web developer feedback. Do we need your LGTM to proceed?

[Mike West]

IMO, this is somewhere on the border between a web-visible experiment and a pure expression of user agent preference regarding flexibility explicitly carved out in a standard.

Rather than debating the feature's philosophical state, I'd simply treat this email as an Intent to Experiment from M104 (current stable) to M107, and give you an explicit LGTM.

Additionally: it would be ideal for the experience you gather in this experiment to fold back into the spec as an "Implementation Consideration" that might help other implementers determine how to use the flexibility the spec provides.

-mike

--
You received this message because you are subscribed to the Google Groups "blink-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email to blink-dev+...@chromium.org.
To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAGD3t5E9r%2BjOcMa5nR3ZgYjNykEyj8bUBmjvszgFYmiBJKP-dA%40mail.gmail.com.

[François Doray]

Thanks! I started the 1% Stable experiment. I will share an overview of the results in ~3 weeks.

[François Doray]

Update: We ended up experimenting with M106+, because there was a bug in the code in prior versions. The M106 Beta experiment has good results. We'll start the 1% Stable experiment this week.

[Mike West]

Hey Jiahe,

Can you summarize what you've learned from the 1% experiment? The design doc isn't exactly clear on what y'all would consider "success" from an experimental perspective, so I'd like to understand what you're evaluating.

I'm also curious about the hop from 1% to 50%. What do you expect to learn from the 50:50 experiment that you're not learning from the 1%? We often do incremental rollouts, ramping from 1% to 5% to 10% and so on, but I think you could do that as part of an Intent to Ship, rather than extending this experiment.

-mike

On Thu, Nov 3, 2022 at 11:57 AM Jiahe Zhang <jiahe...@intel.com> wrote:

We've been experimenting this on 1% Stable on M107 for weeks , and the results are quite encouraging.

So I'd like to request a larger scope of experiments to 50% Stable. Please review.

Best Regards,

Jiahe

[Jiahe Zhang]

We've been experimenting this on 1% Stable on M107 for weeks , and the results are quite encouraging.

So I'd like to request a larger scope of experiments to 50% Stable. Please review.

Best Regards,

Jiahe

On Tuesday, October 4, 2022 at 4:45:42 AM UTC+8 François Doray wrote:

[François Doray]

Hi Mike,

With 1 week of ramped up data of stable, we know that this intervention reduces overall CPU usage from Chrome by a few (single-digit) percents on all desktop platforms. On ChromeOS, this reduces jank (it's typical for CPU-reduction experiments to reduce jank on ChromeOS, due to tight resource constraints). On Mac, it extends battery life. There is no regression for any guiding metric.

We're hoping to experiment on 50% Stable to get more clarity on the battery life impact. It's important to get this data to prioritize future investments. Confidence intervals for battery discharge rate are wide with 7-days of 1% stable experiment. For other CPU reduction experiments, ramping up to 50% Stable was the fastest way to obtain smaller confidence intervals.

Do you recommend sending an "Intent to Ship" asking to ramp up gradually: 1% -> 50% -> 100%?

François

[François Doray]

I no longer think a 50% stable experiment will be required to get small confidence intervals for all our metrics.

I sent an Intent to Ship: https://groups.google.com/a/chromium.org/g/blink-dev/c/5SZB2CFFGqE