After the lightning keynotes and banjo music, I stroll into my first session of the day: MySQL Snapshots and Replication from Oracle by Theo Schlossnagle. As you might imagine, it's about high query counts, lots of data and a limited budget. Starting off with an Oracle cluster, how can we use master-slave replication to handle many distributed queries? He's doing it in the application, selecting individual tables to replicate. It's all quite cute. using triggers to insert changes (inserts, deletes, updates) into a "log table" which is then replayed onto the slaves after a while. For more scaling, he advocated MySQL master-slave replication and LVM snapshots for initial slave startup. Turns out he's explaining what approach works fine for UltraDNS. Jeremy Zawody is sitting behind me and heckling^Wsharing some more information. Unfortunately, Theo often has 30 lines of SQL on the screen, which is quite hard to follow. Best quote: "There are two solutions. One is good, and the other is correct".
