MySQL Geek, self
14 years with Yahoo. The go-to guy for performance issues. Major troller of forums.mysql.com.
You are combining your logs to find out how many _different_ users have connected. This, itself, is somewhat I/O intensive. But then add to that the desire to rollup daily "uniques", into "weekly" uniques. But you find that SUM(), etc, won't work. I will outline a way to store data at less than one bit per unique user, and do rollups.
Performance sags in when the data gets large. This lightning talk discusses 3 classic use cases that suffer this way, then discusses how adding a carefully designed table can take you from meltdown to smooth sailing.
A fast-paced list of one-liners -- MySQL Tips, RoTs, Hints, and Best Practice. These are distilled from watching MySQL users -- from newbies to experts -- successfully (or unsuccessfully) tune, create schemas, write queries, and try to optimize. Some you already know; some you have been wondering about; and some will surprise you. Topics covered: PXC Gotchas, SELECTs-do's & don... [read more]