Matt Mullenweg of Wordpress/Automattic fame is reporting from MySQLCamp talk by Google engineers. The basic ideas behind Google’s use of MySQL for replication are
- using the cheapest hardware possible
- replicating entire databases at a time
- use DNS for load-balancing DB servers
- automated way to bring up a brand new slave
- track the most expensive query
There’s also a talk by Jeff Dean given at the University of Washington last year regarding Google’s BigTable – distributed hash project.
