Performance testing next steps (adoption)

[swissspidy]

Last year we made several big improvements in the performance testing area. To name a few:

• Migrate core performance tests to Playwright
• Added support for taking measurements using web-vitals.js and Lighthouse
  • Started measuring TTFB and LCP
• Extended core performance tests
• Looked into ways to make test results more stable (without much luck)
  • Performance Metrics Stabilization #849
• Made all these utilities reusable
  • Reusable Performance Testing Environment #852
  • Published https://pascalbirchler.com/wordpress-performance-testing/
  • Published https://github.com/swissspidy/wp-performance-action

As a reminder, the two main objectives we had:

1. Improve core/GB performance testing
2. Get other projects to adopt performance testing by making the tooling easier and more reusable

While progress slowed down due to other priorities, there is still a lot of untapped potential this year. I am therefore opening this issue to kickstart a discussion on next steps.

Some enhancements/features that are already on our radar:

• Core
  • Measure performance with a persistent object cache in performance tests
  • Fix the Performance Testing workflow in the 6.4 branch
  • Measure interactivity metrics in performance tests
  • Measure layout stability metrics in performance tests
• Gutenberg
  • Measure more frontend metrics
    • Better demo content, measure Server-Timing, etc.
• Stability
  • Further look into this, try to see if Playground could be an option
  • Maybe build a more powerful dashboard using Grafana or similar
• Adoption
  • GitHub Action
    • Consider moving to the WordPress GitHub org to make it more official
    • Support Blueprints for more realistic setup
  • Maybe provide an all-in-one npm package for running tests & doing comparisons (making it even easier to use)

Some loose ideas:

• Adoption
  • Run performance tests at scale using Tide?
  • Blog post on developer.wordpress.org/news
  • Directly reach out to a few bigger plugins to help them set up performance testing, setting a precedent for others

Curious to hear everyone's thoughts :)

[joemcgill]

Thanks for kicking this discussion off, @swissspidy.

I appreciate the distinction between two main objectives. For now, I'm going to limit my thoughts to the first objective, "Improve Core/GB performance testing".

One of the things that I observed during the 6.5 release is that finding the source of a server timing regression was challenging when that regression was committed to the Core repo as part of a larger Gutenberg sync. I think we can improve this somewhat by updating the performance tests in the Gutenberg repo to include the same server timing metrics that we record for each commit to the WP Core repo (i.e., wpTotal, wpBeforeTemplate, and wpTemplate). While TTFB is a close proximity, the additional noise from network requests and calculating the metric in a headless browser makes pinpointing potential regressions more difficult.

I also strongly agree with your suggestion to improve our demo content, and think this applies as much to the Core tests as the Gutenberg tests. Currently in Core, we are only testing the default homepage for the Twenty Twenty-One and Twenty Twenty-Three themes after importing a Theme Test Data from this commit. While this keeps test content consistent over time, there are a number of limitations to this approach including the fact that some specific use cases that we care about (e.g., measuring the effect of image optimizations on LCP) are not covered by our current test content.

From a visualization point of view, our current dashboards at codevitals.run have become harder to use over time as we've added additional metrics. I'd love to investigate improving or replacing these dashboards with a system where we could more granularly filter results by metric, theme, template, object cache, etc. In doing so, we should also evaluate how we are normalizing and storing the raw data which currently gets normalized before it's saved to the dashboard's database meaning we don't have the ability to build new reports using the original unfiltered data.

Last idea for now, is that when we introduced these tests we used Twenty Twenty-One and Twenty Twenty-Three as representative themes for classic and block themes, but that has proven to be overly simplistic as some performance regressions are only visible based on characteristics of a theme (e.g., how many template part variations they register). At minimum, we should add Twenty Twenty-Four to our test matrix in both repos.

TL;DR:

• Add server timing metrics to the GB performance workflow
• Improve demo content for the performance workflow in both repos
• Add specific use cases (e.g., homepage, page with large LCP image, etc) to our workflows for both repos
• Improve the dashboard for tracking performance over time so it's more user friendly
• Improve data collection process to make raw data queryable in the future
• Add tests for Twenty Twenty-Four

[swissspidy]

That all makes sense to me. I think most of those suggestions are already mentioned in some place or another. I also previously explored a Grafana-based dashboard that would be more user friendly and could be fed raw data.

Some more thoughts on adoption:

Let's start with directly reaching out to some top plugins that we think could benefit from performance testing but don't leverage that yet. Find out why, figure out what's missing, and help them get started.
This way, we can iterate on the tooling before publishing another blog post or improving the GitHub Action.

[swissspidy]

Improve Core/GB performance testing

Mentioned above:

• Improve demo content
  • Needs an owner
  • Priority?
• Add specific use cases (e.g., homepage, page with large LCP image, etc)
  • Also relevant for things like interactivity and layout stability metrics
  • Needs some fleshing out
  • Priority?
• Better dashboard with raw data
  • Needs an owner
  • Priority?
• Add tests for Twenty Twenty-Four
  • This was added in one of the recent refactorings. 2021, 2023, and 2024 are now tested.
• Alignment between core and Gutenberg
  • Core relies on GitHub Actions, while GB uses a sophisticated node script for running tests.
  • Should we have 1 script for both that could also be used by the ecosystem via npm or similar? Or is a GitHub Action good enough?
• Use WordPress Playground for more stability
  • Just for context, this was something that @dmsnell brought up, most recently in this Slack message
  • We could simulate different I/O latencies and intercept every DB and file operation
  • wp-now has some design problems that make it less favorable for testing (needs some refactoring first?)

More thoughts:

• Test more combinations (e.g. Multisite, object cache enabled)
  • Expand performance tests setup wordpress-develop#6515

Adoption

Let's start with directly reaching out to some top plugins that we think could benefit from performance testing but don't leverage that yet. Find out why, figure out what's missing, and help them get started.

The main points I've heard so far in various discussions:

• They use a different testing setup (Codeception, Cypress, Puppeteer, k6) or CI (Bitbucket, GitLab)
• Lack of tests setup in general (e.g. not even unit tests in place)
• Focus is more on backend performance rather than frontend performance
• Lack of resources for implementing additional kinds of testing
• No testing strategy, e.g. even if they had the whole setup, they wouldn't know where to start
• Metrics stability concerns
  • Makes it especially difficult to tie metrics change to specific code change

Mentioned above:

• GitHub Action
  • Great, but lack of setup step for plugins
    • Add blueprint support so that plugins can easily prepare their testing environment
      • Blueprint support swissspidy/wp-performance-action#71
      • @adamziel Curious to hear your thoughts on that one
    • Concern: wp-now appears to be unsuitable for testing like this :( (see above)
  • Once this is in place we can think about further blog posts
• All-in-one npm package for running tests & doing comparisons
  • Does it make sense? Usually not advisable to run tests on your own machine vs. a dedicate one

[adamziel]

I have a small backlog but I just added this to our board to make sure this gets a follow-up.

[joemcgill]

Thanks for the updates, @swissspidy. This looks like a good list to me. I do think that separating the Core/GB work and the Adoption work into two separate epics likely makes sense, and will make it easier to track progress of the two initiatives, even if there is some overlap.

From my perspective, focusing on improving and aligning the metrics that are taken and the way dashboard to display them between the WP and GB repos should be a high priority, so we're able to more quickly identify the likely cause of performance regressions based on server timing data earlier in feature development in the GB repo. We also need to fix some existing issues with the WP Performance workflows that have caused big jumps to show up in the dashboard due to the way the baseline checks were taken and add metrics for the TT4 theme in our WP dashboard view.

[adamziel]

Great, but lack of setup step for plugins
Add blueprint support so that plugins can easily prepare their testing environment
swissspidy/wp-performance-action#71
@adamziel Curious to hear your thoughts on that one

@swissspidy do you mean preparing a generic GitHub action that plugin authors could use to get a performance testing link on their PRs? If yes, that sounds fantastic. There's a related Playground issue and here's some inspiration to reuse:

• Auto-post a Playground preview link on every PR gutenberg#60795
• https://github.com/Automattic/themes/blob/trunk/.github/workflows/preview-theme.yml
• https://github.com/WordPress/blueprints/blob/trunk/.github/workflows/preview-comment.yml

If, on the other hand, that's about running E2E tests locally or in CI, @WunderBart's explorations here are highly relevant:

• PoC: Run E2E tests with WP Playground gutenberg#62692
• PoC: Visual regression testing with Playground gutenberg#62729

If I misunderstood, would you share some more details about that question?

[swissspidy]

@adamziel The idea is to provide a generic GitHub Action (https://github.com/swissspidy/wp-performance-action) that automatically runs e2e performance tests in CI. It already works (using wp-env) but blueprint support would be a killer feature. This way, projects can declaratively add test data/fixtures.

@dmsnell originally suggested using Playground as we could then have full I/O and DB control, which would hopefully help with getting stable results and would be a nice side effect.

WordPress/gutenberg#62692 comes close, it's just not stable enough :-)
Previously I used wp-now to avoid building Playground myself, but using Playground directly of course also works.

Regarding wp-now, Dennis mentioned some architectural issues that "make it less favorable for testing"

I also saw https://github.com/adamziel/wordpress-php-blueprints/ but not exactly sure what it does 🤔 Could I leverage this to use blueprints with an arbitrary setup? This way, I could already use blueprints while Playground is still being improved upon.

[swissspidy]

@joemcgill Extracted core improvements to #1380 now, we can keep this one for the adoption part

[swissspidy]

Here's a first successful attempt at using Playground for the testing environment: swissspidy/wp-performance-action#173