This page provides query throughput information on a per-queue basis, for worker node queries only.
This information provides the capability to rapidly determine if WLM configuration changes have reduced or increased query throughput.
This is a critical question; if we have a Redshift cluster, and users are complaining there is a large backlog of queries, and we say double the number of slots, then twice as many queries can execute in parallel, but they each have half the memory available to them.
It is not obvious whether this change will help, or hinder - ti could be the loss of memory makes queries so very much slower that the increase in query parallelism is outweighed by the large increase in the time taken for queries to complete.
There is obviously a need for a measure of query throughput, which is provided by this page.
However, the definition of throughput is important and must be understood to understand what you’re looking at in this page, and actually is quite a subtle and complex matter.