1 We use MySQL, but mostly as a key-value store that is partitioned horizontally with our own sharding scheme in a service atop MySQL.
2 For example, we have tables that map usernames to user IDs, and those can't be sharded based on user ID.
3 For the majority of our data we shard based on user ID*, all tables exist on all shards, and the schemas of our tables look like:

트레이시 츄는 쿼라 있다가 핀터레스트로 옮겼구나. 예뻐서 더 기억나는

트위터 반응

@lswdev: 핀터레스트 DB구조: Pinterest: What is the structure of Pinterest's database? - Quora http://t.co/mxy2eEwglF