[META] Dashboard Performance

[thomasneirynck]

This meta-issue collects issues around Dashboard Performance.

Efficient data fetching

Dashboards should request the data for the widgets in the most efficient way.

• [Research] bsearch investigation #166206
• Global time range on Dashboards should be rounded #166210 / Round date by default in date picker #94280 :
• [Analytics] Allow opt-in for random_sampler agg #132966
• [Dashboard] Field-caps requests are made for unused default dataviews #162403
• [data views] async and partial loading of field list #152159 / [Lens] Should only load necessary fields from DataViews when viewing charts/maps #166213
• Allow resolution of Data View without resolving all fields #139340
• [Dashboards] Enable easy filtering by data tier #166948
• [Research] Data format improvements for charting (arrow) #175695
• [Research] Kibana Expression Language impact to Lens charts performance #175700
• [Research] Query consolidation on Dashboards #175702
• Disable Refresh while not in focus #1878
• [Serverless] [Unified Search] Increase minimum refresh interval to 5s. #176256

Efficient code loading

Load code most efficiently. Improve packaging.

• [Meta] Reducing Kibana bundle sizes #88678

Efficient DOM rendering

Dashboards should efficiently manipulate the DOM to avoid FOUC.

• [Dashboards] Dashboards should only have single React render tree #166214

Perceived performance

Improve perceived performance by UX-flow optimizations

• [Dashboards] Collapsible Panels #1547
• [Partial Results] Incremental results visualizations UX #93007
• [Lens] Inline editing of lens panels on a dashboard or canvas #166833
• [Maps] inline editing of maps in Dashboards #166950

Debugging

• [Dashboard] Provide dashboard queries API #175487

[elasticmachine]

Pinging @elastic/kibana-presentation (Team:Presentation)

[Dosant]

Wanted to share some random observations and ideas that I didn't noticed in the list so far, maybe some of those deserve a closer look:

Some larger items:

• React 18 has new APIs for efficient prioritizing of updates https://react.dev/reference/react/useTransition. I wonder if using this new feature might help us with perceived performance in some places. Also, I wonder if updating to React 18 would increase performance by itself. Not sure how much effort the update is and if it was tried before.
• There is an existing deferBelowFold experiment, I saw it is used on large internal dashboards, and it seemed very useful. Not sure why we plan to remove it [Dashboard] Remove defer below fold setting #176535 instead of investing more into it. Maybe worth another look

Tiny quick win:

• It seems like we're still using some resize observable polyfill instead of native implementation. Removing it should cut a couple of milliseconds.
• Looking at the profile there are isExpressionValueError and getType from the expression layer that stand out with their self-time. Maybe there is a quick small win with simplifying isExpressionValueError because of how often it is called.

[thomasneirynck]

Closing this meta-ticket. Items can be tracked in the project https://github.com/orgs/elastic/projects/1307