but in all seriousness, it depends. adding an index. removing/consolidating indexes. breaking the queries down into individual UoW and forcing intermediate materialization. identifying platform-specific optimization barriers. rearranging sufficiently complex query semantics to force behavior you expect.
99% of the cases i’ve personally had to resolve over the last 15 years have been the result of sql hero queries that try to do everything all at once. this is exacerbated by orms that generated bad sql but was acceptable at low cardinalities. under scaled-up concurrency and data volumes they can’t deliver the necessary performance anymore.
simonbarker87|1 year ago
“When you’re fed up of keeping up just retire in to SQL, it’s the best pension there is”
Vt71fcAqt7|1 year ago
swasheck|1 year ago
but in all seriousness, it depends. adding an index. removing/consolidating indexes. breaking the queries down into individual UoW and forcing intermediate materialization. identifying platform-specific optimization barriers. rearranging sufficiently complex query semantics to force behavior you expect.
99% of the cases i’ve personally had to resolve over the last 15 years have been the result of sql hero queries that try to do everything all at once. this is exacerbated by orms that generated bad sql but was acceptable at low cardinalities. under scaled-up concurrency and data volumes they can’t deliver the necessary performance anymore.
chadcmulligan|1 year ago