Peter Boros

When I talk or write about benchmarking, a repeated question I get asked is how to make the graphs. For some of those graphs, check the links below.

https://archive.fosdem.org/2014/schedule/event/lepeterborosrulez/ http://www.percona.com/blog/2014/08/12/benchmarking-exflash-with-sysbench-fileio/

In this talk, I show how to make those graphs with R and ggplot2, and how to prepare the benchmarking data for that.

This session will be about benchmarking MySQL and disk IO subsystems with sysbench and interpreting the results. In our consulting company, I helped a reasonable number of customers with sysbench so I know the common caveats most people run into. This talk will cover benchmarking IO subsystems with fileio tests, as well as benchmarking MySQL.

Many applications are using MySQL’s built in replication to provide high availability with either manual or automated failover. One issue with this is that after failover, the former slave's buffer pool may be cold, and it becomes IO bound.

This talk is about how we solved this issue and provided fast failover for a high traffic site using Percona Playback to replay the database traffic continuously on the failover hosts.

This talk is about the solutions we used there, and the measurements we performed to choose the appropriate solution for the issue.