大多数产品的规模永远不会触及需要专用数据库的瓶颈,先从简单方案出发,等真正遇到功能需求时再升级。

原文链接

我的结论

关键信息

需要升级到完整数据库的信号:

只要不触碰这些边界,SQLite 或更简单的方案就够用,不要为了“以防万一”引入过重的基础设施。

摘录

Instagram was at 400 million daily active users and still running PostgreSQL as their primary data store (Instagram Engineering). Most products never get there.

And when you do, SQLite running from a flat file handles 90 million daily active users on a single server.

You need to query by more than one field.

You need joins.

Multiple processes need to write at the same time.

You need atomic writes across entities.

我的补充

ref