Page MenuHomePhabricator

[Investigate 4h] will instantCommon overload commons?
Open, Needs TriagePublic

Description

We want to enable InstantCommons on cloud but are not sure how much traffic this will result in for real commons. We want to make sure we're being good internet citizens.

Questions we want answered:

  • Does it cache:
      • The actual image files on the server?
        • Check this by turning on InstantCommons locally + adding loads of files + using the browser network tools to see if they come directly from commons or if they come from the local wiki.
      • Metadata i.e. the links to those files?
    • check this by looking in the cache and seeing whether stuff is cached.

If so:

  • where are the cashed files stored?
  • will we get rate limited?
    • this is hard to be certain about but maybe try adding loads (50+) of files and see what happens locally

Event Timeline

I just hope that the test goes as planned-

We can also use $wgForeignFileRepos to customizes caches or uses QuickInstantCommons extension.

Macro challenge-accepted:

Per commit for T114098:

* $wgUseInstantCommons now hotlinks Commons images by default instead of downloading originals and thumbnailing them locally. This allows wikis to save on CPU and bandwidth while reducing time to first byte for pages, even without a thumbnail handler. See $wgForeignFileRepos documentation for tweaks.

Commons thumbnails are served on Wikipedia, so I'd be surprised if the remote use of Commons thumbnails by what's still a relatively small project imposed significant load, especially if the images are only accessed on view.

That said, maybe latency will be an issue, since only caches are available in Europe? I'm not clear why InstantCommons was so bad in the QuickInstantCommons performance test.

Tarrow renamed this task from [Investigate] will instantCommon overload commons? to [Investigate 4h] will instantCommon overload commons?.Thu, Apr 18, 1:23 PM