Stop Indexing the Query
summary
Methodology scores queries by an inefficiency score = (logical + physical reads) / rows × execution count, then groups statements by table and ranks them, keeping only those with reads to rows > 8. Uses DMVs like dm_exec_query_stats and plan XML to spot missing index recommendations, producing vs consuming operators, implicit conversions and parallelism. Checks data churn (over 50% per day raises caution), removes duplicate/unused indexes, measures post-change reads and execution cost; approach reduces aggregate IO and cloud compute spend.
AI-written summary · about
chapters (10)
- 0:00Holistic workload approach
- 1:31Indexing methodology
- 3:53Defining inefficiency
- 7:03Analyzing query performance
- 11:17Data-driven prioritization
- 14:09Inspecting query plans
- 19:07Index validation and cleanup
- 23:29Managing data churn
- 25:52Iterative optimization
- 30:28Cloud, cost, and conclusion
related talks
41mStop Indexing the Query
53mIndexing Internals for Developers & DBAs- 53mIndexing Internals for Developers & DBAs
07mIndeksimi dhe përmirësimi i performancës në SQL
52mPostgreSQL Index Strategy Guide
01mHow well do you know your MySQL Indexes?
19mStabilizing Performance with Query Store
46mA Rocket Scientist Perspective on Index Column Selection