A Rocket Scientist Perspective on Index Column Selection
summary
Walks through hands-on demos against AdventureWorks2017 big using provided scripts and execution plans to prove index column order matters: a 26 MB nonclustered covering index drops a query from ~29,295 logical reads to ~264 (sometimes to 4–5 reads) by ordering status before territory, and reversing order raises reads and cost. Shows covered indexes let UPDATEs use index seeks instead of clustered scans, and that an unused secondary index can force worse plans.
AI-written summary · about
chapters (5)
- 0:00Start
- 3:57Overview
- 35:33Covered Indexes for Update Statements
- 41:30Create a Covered Index
- 44:58Conclusions
related talks
53mIndexing Internals for Developers & DBAs
53mIndexing Internals for Developers & DBAs
53mIndexing Internals for Developers & DBAs
07mIndeksimi dhe përmirësimi i performancës në SQL
33mDriver for Data-Driven Change
53mSQL Server Internals: The Practical Angle
41mStop Indexing the Query
19mStabilizing Performance with Query Store
