When you execute your first query, Varada begins learning your workload and accelerates the datasets accordingly, so that upcoming queries are accelerated.
To see the query acceleration details go to the Query Steam and select the query line. In this example, you can see that 100% of the query was accelerated by Varada.
The predefined queries include queries with various levels of selectivity, such as a full scan query with no filters, a cohort analysis query, a selective query, and a query with a selective join. Notice that the more selective the query, the greater the acceleration.
You should expect an acceleration of up to X40 for this set of queries when comparing the performance of the Varada catalog to that of the Hive catalog used by the Vanilla Trino Query Engine.
You can toggle between the catalogs to test performance.
The following syntax examples assume that you created the tables under the trips_data_sample schema.
To run the queries on the Hive catalog, first run:
To run the queries on the Varada catalog, first run:
Updated 25 days ago