Before twiddling the engine parameters start by looking at the actual queries and query plans they generate. A few bad ones will eat up far more resources than you'll ever get back from tuning.
Rapping otters and automated database knob-twiddling: An obvious combination in some universe or other
A university spin-out startup has announced a private beta of an automated database tuning service which its founder claims can double the performance or halve the cost of the popular AWS Relational Database Service. Among its marketing hype, though, is the, erm, novel approach of launching a hip-hop album of beats and …
Friday 14th May 2021 14:59 GMT yetanotheraoc
"A few bad ones will eat up far more resources than you'll ever get back from tuning."
Apart from, as mentioned in the article, some places don't want you knowing about their queries; there is also the fact that poor performance could be either the overall database or the individual queries. So if you tune the database in some reliable way, and the application still performs poorly, then you know it's the queries. Of course it almost always is the queries, but the same people who are resistant to writing good queries are incredibly resistant to making bad queries better. One team I was on, the most experienced guy always copy/pasted the *exact same* hints at the top of *every one* of his select statements -- even across databases. I think he just wanted management to believe he had done some tuning. At least after applying the OtterMation management knows where the problem lies.
Saturday 15th May 2021 04:29 GMT Gene Cash