Day 0 of the MySQL Conference was a day unlike any other day. It was, in fact, tutorial day. While regular days of the Percona Live MySQL Conference feature 50 minute sessions, usually split into a 40 minute talk and a 5-10 minute question period, tutorials are 3-hour-long sessions (with a generous 10 minute break in the middle for those that wish to go to the WC) that provide an in-depth dive into some aspect of MySQL. Due to the length of the tutorials, they are more in-depth and technical than individual sessions can be, but at the same time, we are limited to 2 tutorials slots per day instead of the 5 for sessions.
The tutorial schedule for the conference is located here, and with so many good ones, it was hard to choose which one(s) to go to.
For the morning session, I attended Peter Zaitev’s tutorial entitled “InnoDB and XtraDB Architecture and Performance Optimization”. As most of us know, Peter Zaitev is the CEO and Co-Founder of Percona and an expert in InnoDB/XtraDB architecture, performance, and optimization. The session was quite good and informative. Peter went over the core InnoDB/XtraDB options in regards to performance and more. It was interesting to learn what are some of the more exotic options available in XtraDB (Percona’s variant of InnoDB with performance improvements), such as the ALL_O_DIRECT option for innodb_flush_method, and how they impact performance. Considering that the vast majority of the deployments of MySQL in businesses use InnoDB, the talk was quite relevant and informative.
In the afternoon, I attended “Linux and H/W optimizations for MySQL” by Yoshinori Matsunobu of Facebook. If I had to say one thing about this session, it would be that it was one giant info dump and I need another 30 hours to study the numbers. Yoshinori compared MySQL performance across various hardware configurations, software configuration, memory settings, versions of MySQL, MySQL configurations, etc. and gathered some interesting and sometime unintuitive numbers. He also showed how modern hardware amplified the replication lag issue, which is common to slaves in MySQL, and how to deal with it by judiciously using SSDs.
Leave a Reply