Let’s talk about the trade-offs.
Auto-updated stats are triggered by queries needing to access a table with “stale” stats. (It doesn’t matter how we define stale, so lets not think about it.) With the exception of very small tables, auto-updated stats use a random small data sample of the table data; in contrast scheduled jobs that usually have a larger sample size defined—I always say “
FULLSCANif you can!”
Read on for Andy’s thesis.