Re:Why we're helping more wikis move away from Fandom
by bawolff on Thursday October 10, 2024 @02:28PM
(#41801765)
Varnish caching really only helps if the majority of your traffic is logged out requests. Its the sort of thing that is really useful at a high scale but matters much less at a low scale.
Application level caching (memcached/redis/apcu) is super important even at a small scale.
Most of the time (unless complex extensions are involved or your wiki pages are very simple) mediawiki should be io-bound on converting wikitext -> html (which is why caching that process is important). Normally if db is healthy, db requests shouldn't be the bottle neck (unless you have extensions like smw or cargo installed)
Re:Why we're helping more wikis move away from Fandom
by jchw on Thursday October 10, 2024 @03:20PM
(#41802448)
Most of MediaWiki seems to avoid too much trouble with contention in the database, but I was seeing it prior to enabling application-level caching. It seemed to be a combination of factors primarily driven by expensive tasks in the background. Particularly complex pages can cause some of those background tasks to become rather explosive.
The calculus on this probably changes dramatically as the RPS scales up, though. Not doing work will always be better than doing work in the long run. It's just that it's a memory/time trade-off and I wouldn't take it for granted that it always gives you the most cost-effective end result.