Wikipedia uses MySQL and although I worked there for 7 years I never did much work directly with the DBs.

It's been 9 years since I worked at deviantart.com but last I knew the site still operated on a large sharded MySQL cluster divided into something like 32 individual shards where each user's data was on a shard determined by their userid.

Essentially just `shard = userID % shardCount`;

The application would connect to all of the servers and for many queries you have to send the query to all 32 servers, or some subset based on the records you seek, and then merge the results of all of those queries.