[HOLD for payment 2024-10-30] [HOLD for payment 2024-10-29] Benchmark performance bottlenecks when opening chats

[adhorodyski]

Coming from the discussion thread under https://expensify.slack.com/archives/C05LX9D6E07/p1728314452317409

Problem

Lack of solid performance benchmarks across the core app workflows, baselines for proposing performance improvements.

Solution

• Focus on 1 scenario of opening a chat
• Benchmark it in detail to get the granular results (also for each phase) so we have solid numbers for whom specifically it's slow and why. There are also analytics we can use for this to have more datapoints.
• Pinpoint the areas for improvement we see
• Map how the proposal from initial thread/nitro modules/anything other can fit to improve it

cc @hannojg

[adhorodyski]

@mountiny could you please assign this one to me & @hannojg? Thanks!

[hannojg]

Thanks! 👋

[adhorodyski]

Update
Our bottom line is to get super solid on what precisely is slow (have granular measurements) and for whom specifically.

Who
To be able to say how many people something affects (in different percentiles), we need to have something to segment by. We'll extend our performance analytics' metadata (in all envs, including e2e tests) to include the below fields. These will help us group measurements by the types of accounts. (PR: #50881)

• number of reports ✅
• number of reportActions 🔄
• number of personalDetails ✅
• number of policies 🔄
• number of transactions 🔄
• number of transactionViolations 🔄

What
To be able to pinpoint what exactly is slow, we'll implement granular metrics in code, across both App & Onyx. Here's a breakdown of all the phases we decided to focus on.

• APPLYING_UPDATES (coming over https, until we schedule notifying subscribers)
  • DATABASE_WRITE (actual save operations like multiset and multimerge)
• NOTIFYING_SUBSCRIBERS (multiple key(s)Changed triggered by scheduleNotifyCollectionSubscribers)
  • DATABASE_READ
  • REACT_PROVIDER - (useOnyx/withOnyx)
• ONYX_DATA_POSTPROCESSING (from receiving a new update (subscriber got called) until we have the final data)

How
TBC

[adhorodyski]

cc @mountiny we'd like to ask you for any feedback about this 1st batch (please invite others!). Is there an area we're missing? I want to make sure we focus on the right parts when profiling this, the exact tools (like e2e suites) will come next once we fully agree on the who&what.

[melvin-bot]

Reviewing label has been removed, please complete the "BugZero Checklist".

[melvin-bot]

The solution for this issue has been 🚀 deployed to production 🚀 in version 9.0.51-4 and is now subject to a 7-day regression period 📆. Here is the list of pull requests that resolve this issue:

• Extend perf analytics metadata #50881

If no regressions arise, payment will be issued on 2024-10-29. 🎊

For reference, here are some details about the assignees on this issue:

• @hannojg does not require payment (Contractor)
• @adhorodyski does not require payment (Contractor)

[melvin-bot]

⚠️ Looks like this issue was linked to a Deploy Blocker here

If you are the assigned CME please investigate whether the linked PR caused a regression and leave a comment with the results.

If a regression has occurred and you are the assigned CM follow the instructions here.

If this regression could have been avoided please consider also proposing a recommendation to the PR checklist so that we can avoid it in the future.

[adhorodyski]

The extended analytics setup is merged now:

• to be confirmed it's available in Firebase for newly reported events
• to be merged into local-based profiling tools (on top of react-native-performance within the Performance module?) so we can use the same parameters in eg. e2e workflows cc @hannojg, I think this is the last thing we have to do when it comes to the 'segmentation' part

[adhorodyski]

• Next part is to implement the granular metrics as listed above

also cc @hannojg, do you think these should also be reported to Firebase OR we just focus on being able to profile this behind some flag?

[adhorodyski]

I'm back to this issue after some inactivity from my end, will post an update tomorrow :)

[adhorodyski]

For starters I'm investigating why updated metadata fields are not visible in Firebase. They're in code here for 2 weeks now.

cc @mountiny what is the last tag available in Hybrid, maybe it's a too short time window? #50881 (comment)

[adhorodyski]

Besides above I'm working on putting code markers in the correct places now.

[mountiny]

@adhorodyski are you getting staging data there too? latest staging is 9.0.55-9

[melvin-bot]

Skipping the payment summary for this issue since all the assignees are employees or vendors. If this is incorrect, please manually add the payment summary SO.

[hannojg]

Sorry, this completely slipped through the cracks!

do you think these [the granular metrics as listed above] should also be reported to Firebase OR we just focus on being able to profile this behind some flag?

I think behind some flags so we can test those locally. I think it might be a bit too much to track all those events and send them to analytics. Lets get those implemented! How can i help on those?

[adhorodyski]

I'm starting with the more high-level marks here, but for this very benchmark we need Performance.markStart/End() calls as described above so we can collect them to memory. Can you possibly pick even a few of them @hannojg?

[hannojg]

Yes, will check! This will also involve making a few changes in onyx to enable these kind of measurements!

[hannojg]

Working on a PR for onyx over here:

• chore: re-enable performance metrics react-native-onyx#598

expecting to wrap it up by tomorrow!

With these changes we should be able to record all steps of what onyx does using the performance API / react-native-performance. We can then use the same API in the app to create reports for different actions performed.

[adhorodyski]

Reviewing now!

[adhorodyski]

Approved the PR to Onyx. I'll start adding the other metrics' calls (app-specific) into the app next week.