I have a query that does multiple joins on several tables and uses a covering index. It seems to run a lot faster when executed directly from the mysql cli instead of from
There's a persistent connection in Workbench, so there's no overhead in that regard. Executing a query should be equally fast in both CLI as well as Workbench, however data transfer is a bit slower sometimes because Workbench locally caches results first to be able to sort over it when the user clicks a header field in the resultset view (including multi column sort). Try repeating the query in both tools and see if the times still differ. The first run of a query is usually slower than following runs due to the execution cache.