Trino parallel query execution time increases with users

Trino
1

When I fire a select query on a single partition with a simple where clause the query execution time is 6-8 secs now when same query is fired by multiple users parallely the execution time for each query goes up to 14-15 Secs, now same query for 5 parallel uses it goes to 24-26 Secs.

I’ve tried to enable bloom filters but currently it is only available with ORC file format not with parquet.

Is there a way to optimize my trino cluster for parallel queries ?

Trino Cluster Configurations:
1 Master, 4 workers, file type parquet, partition enabled

2

Trino is already optimized for multi-user query processing. Of course your overall connection to the object storage and your cluster power still have an impact. In your case it might be pinned due to all being a single partition. You will have to look at the explain plan for the queries to diagnose more and provide more details.

I would also just use at your actual workload rather than testing some arbitrary scenario that might not reflect a real usage.