High performance replication

Replication and High Availability Strategies
23 April 4:50pm - 5:40pm @ Ballroom B

Facebook social graph persists in multiple datacenters - and replicating it all around has been a performance challenge and bottleneck.
This talk covers field experiences of having high throughput replication requirements while serving massive scale and load environment.

From various naive prefetching methods to:

  • Faker & InnoDB Fake Changes
  • Multi-instance replication
  • Binary log transfer optimizations
  • Active replication performance management


Domas Mituzas
Database Engineer, Facebook
Domas is a member of Facebook database performance team. In past he volunteered for Wikipedia and worked at Sun Microsystems and MySQL.