Hello Sam, thank you for the kind words.
And also thank you to Chris and Martin for yours.
With regards to the “optimizations” mentioned in the product update: usage change over the course of 2022 wasn’t really the motivating factor here (given it’s only been one year so we’re fully expecting fluctuations), It’s more about trying to increase the system’s efficiency and limit areas where things could break at scale - in preparation for future demand. There has been some increase in costs over the 2022 calendar year, as shown in the finance report’s bar-graph and associated paragraph, but that was primarily due to staffing changes, not hosting costs.
Our initial system architecture focused heavily on Wikipedia database dumps and now we’re focusing on expanding the “realtime” data feeds. We decided to rebuild the architecture to look more like an event-based system, to work alongside exported-data products, instead of adding on more services on top of the dump generation.
As mentioned in the product update, for those interested in this topic there are quarterly summaries of the technical work on the project’s MediaWiki page: https://www.mediawiki.org/wiki/Wikimedia_Enterprise#Updates But, if you’d like to talk through the implementation details and the thinking behind it, then the best format is a real-time conversation. I encourage you to come to the “office hours” public meeting this Friday - etails at the top of the project page on Meta. Even if you can’t make it to that, we’ll make sure to discuss the product update and I’ll be uploading the recording afterwards to Commons, with a timecoded summary of the conversation.
Sincerely,
Liam
