SQL Server 2019 Aggregate Splitting
Exploring a new query optimizer exploration rule in SQL Server 2019 can split large batch mode hash aggregates to avoid or reduce spilling to tempdb.
Articles
Page 3 of 10
— showing articles
37 to
54
of 177
A bug with Halloween Protection and the OUTPUT Clause
SQL Server has a bug with Halloween Protection when using an OUTPUT clause on a DELETE statement.
How MAXDOP Really Works
A MAXDOP n hint does not limit a SQL Server execution plan to using a maximum of n CPU cores concurrently 🤯
Pulling Group By Above a Join
How and when the SQL Server query optimizer can pull a logical Group By operator above a logical Join.
New Metadata-Only Column Changes in SQL Server 2016
New instant column data type changes possible in SQL Server 2016 when using data compression.
Finding Distinct Values Quickly
A few different ways to retrieve distinct values from a table, including some big "it just runs faster" wins in SQL Server 2019.
The Eager Index Spool and The Optimizer
The SQL Server Eager Index Spool execution plan operator, and the optimizer rules capable of adding it into plans.
Nested Loops Joins and Performance Spools
Performance spools and how they can be used by the optimizer to reduce the estimated cost of nested loops joins.
Batch Mode Normalization and Performance
The internals of SQL Server batch mode data normalization and how it can affect execution plans and performance.
Batch Mode Bitmap Demos
Demo scripts for SQL Server batch mode hash bitmap selectivity estimates. Simple and complex bitmap types. Compiled and compressed bitmaps.
Batch Mode Bitmaps in SQL Server
A deep dive into SQL Server batch mode bitmaps revealing undocumented details that can greatly impact performance.
Some ANY Aggregate Transformations are Broken
Two cases where the introduction of the internal ANY aggregate by the optimizer may cause wrong results.
Apply versus Nested Loops Join
Understanding the differences between SQL Server APPLY and nested loops joins, including details of query optimizer transformations.
Minimal Logging with INSERT…SELECT and Fast Load Context
Minimal logging for inserts into b-tree indexes (empty or not) using the Fast Load Context, and why it might not always work as expected.
Minimal Logging with INSERT…SELECT into Empty Clustered Tables
The factors SQL Server considers when enabling minimal logging for inserts into an empty clustered table.
Minimal Logging with INSERT…SELECT into Heap Tables
The factors SQL Server considers when enabling minimal logging for inserts into a table without a clustered index.
Columnstore Grouped Aggregate Pushdown
A deep dive into columnstore grouped aggregation pushdown, showing where this optimization is unsuccessful or not attempted, using the undocumented SQL Server command DBCC CSINDEX.
Why IGNORE_DUP_KEY is slower on clustered indexes
Why the IGNORE_DUP_KEY index option is much slower on a clustered index than a nonclustered unique index. Detail execution plan analysis and internal engine details.