Query performance is a big part of how customers experience their interaction with data. Looker provides many ways in its modeling layer to optimize for performance. The key to understanding the Looker approach is to first understand where the difference comes from:
- rewrite the individual query
- re-tool the db to handle better some of the queries as the issue presents itself
- design of how all of the queries will be structured
Basically, as LookML analysts, we have much more context to:
- how all the pieces relate to each other in our dataset
- what queries are important/frequent and should be a priority
Because we have more context, we can much better anticipate the needed performance changes depending on the type of dataset. Below is the presentation outlining 3 types of common dataset patterns and how recognizing those patterns helped resolve performance issues.