Performance practices for minimizing replication delay

For many applications, significant (seconds to minutes) replication delay is not acceptable. In this session, the speaker will cover practices to avoid or minimize replication delay. The following topics will be covered. - How to check replication delay correctly Using "replication clock table" is a common practice. Seconds_Behind_Master often tells lies. Probably the most annoying case is Seconds_Behind_Master returns 0 and SQL/IO_Running return Yes even though I/O thread is not actually running. The speaker will show some interesting examples. - H/W and my.cnf configurations Using larger memory or using SSD is very effective. When using SSD, reducing write volumes (i.e. disabling doublewrite) is also effective. The speaker will show some examples how these optimizations are effective. - Slave Prefetching Slave prefetching is often very effective to reduce replication delay. The speaker will introduce "Replication Booster (http://yoshinorimatsunobu.blogspot.com/2011/10/making-slave-pre-fetching...)" here. - Multiple SQL threads MySQL 5.6 supports parallel SQL threads. The speaker will show some benchmarks how parallel threads helps. - Case study The speaker will show some examples of replication delay, and will explain how to solve them. (i.e. bad queries, massive random reads, sudden stalls, etc)