SQL Server 2019 CTP3.1 launched an optimization to deal with last-page insert rivalry. This takes the type of an index choice referred to as
One imagines this could possibly be an adaptation of Bw-Tree or Bz-Tree. Nevertheless, these depend on variable-sized pages whereas the present storage engine requires fixed-size pages.
How is the optimization carried out? How are present B-Tree algorithms altered by this optimization? In what circumstances would I select not to deploy this selection?
A patent for a reverse key method.
I had a fast look utilizing DBCC PAGE, evaluating 2017 to 2019 and 2019 with and with out OPTIMIZE_FOR_SEQUENTIAL_KEY on an distinctive clustered index for an int IDENTITY column. There was nothing that clearly defined the brand new behaviour. This makes me assume its an algorithmic factor, slightly than a structural factor, which is sensible.
A weblog submit from MS.
This characteristic appears to centre on the detection and avoidance of convoys.