Details
-
Bug
-
Status: Resolved
-
Major
-
Resolution: Invalid
-
None
-
None
-
None
Description
Currently Phoenix early-starts TableResultIterators.
I do not think that is correct; what happens is that the scanner will start working simply by being created, and the work is not handled by the threadpool. In other words merely creating the iterator will start to do work, regardless of how large the Phoenix pool is size. It might seem to be faster, but only because we're throwing more threads at it than we say we would. Making the threadpool larger should get us back to the same the performance.
The main task here is to perf test this.