Ten Rules for Doing a PL/SQL Performance Experiment
Ive all too often been asked to comment on a speed difference between two PL/SQL tests where each is presented to me as just SQL*Plus spool files (prompts and all). Sometimes this is copied and pasted into the text of an email. Often, the text of the second spool file differs from the text of the first in just a couple of tiny details. Worst of all, the times are measured, for just a single test run, using SET TIMING ON. And the times are shown, like Elapsed: 00:00:12.36, with indefensible precision. Sometimes the times are shorter. This is no way to conduct a performance experiment! This session tells you how to do your experiment properly with reference to a real, and interesting, case study. I said PL/SQL in the title, because thats what Im known for. But really, Im talking about any single-user test that runs entirely inside Oracle Database. And I said ten rules because a talk has to have a catchy title. Of course there are many more rues than just ten.