Optimization of Insert Stored Proc - Parallelization Idea

[sfelman]

The Insert Algorithm could be run with some parallelization.
The main insight is that you can only have 1 insert running at each level of the tree at 1 time.
Once the prior insert has reached the point where it has compared to the parent node, you can start the next one and you can then do your comparisons in parallel. Up to 1 comparison per level of the tree at a time. As the tree gets larger this performance increase could be significant.

The Insert stored procedure would need to take a table parameter rather than a single word and there would need to be some way to track which word is next to insert, the current parent node Id for each word currently being inserted, and which are done inserting.

[sfelman]

This idea could still be added for a bit more parallelization and optimization of the insert.