Wikimedia Performance

Wikimedia Performance provides tools to analyze and improve the site performance of Wikipedia.

We do this through a combination of live instrumentation, continuous monitoring, and debug profiling.

Frontend

• Navigation Timing

  We collect real-user monitoring (RUM) metrics from page views such as page load time, network latency, and first paint. Sliced by region and device type.

• Google Web Vitals

  RUM metrics collected by us and categorized using Google's thresholds for "Good", "Need improvement", and "Poor".

• CPU Benchmark

  Shine a light on what kind of devices Wikpedia readers use. Compare device speed of mobile phones and desktop between different countries.

• Synthetic tests

  Monitor frontend performance from a controlled lab environment using WebPageReplay and Browsertime. Provides a high level of detail in recording and profiling.

• CruX Report

  Keep track of how Wikipedia is doing from the point of view of Google Search, through metrics Google collects from Chrome users.

• AS Report

  Insights into the effective performance of Internet providers around the world.

You can find a full list of dashboards in Grafana that plot other aspects of frontend performance.

Backend

• Flame Graphs

  Aggregate flame graphs of sampled stack traces from live MediaWiki requests. Reported daily and hourly, sliced by entry point.

• Request profiling

  High-resolution flame graphs on a per-request basis with Excimer UI. Accessible via the WikimediaDebug browser extension.

• XHGui profiling

  Analyze a given request with XHGui to obtain an accurate call tree, call count, and memory usage. Accessible via the WikimediaDebug browser extension.

• Pageview Timing

  Backend latency from the MediaWiki application when responding to page views.

• Save Timing

  The overall time to process an edit in MediaWiki, with a break down by entry point, page type, and user type.

Refer to Wikitech for more tools and lower-level statistics of backend application performance.