Paul White explains why the IGNORE_DUP_KEY index option on a clustered index can result in much poorer performance than for a nonclustered unique index.
Jonathan Kehayias shares insight about new UDF information in showplan, and shows that only some tools will show it (SSMS 17+ & SentryOne Plan Explorer).
Continuing his series on optimization thresholds, Itzik Ben-Gan dives into the calculations behind stream aggregate costing when the data needs to be sorted first.