Perfomance in create table

[ericfreitas3]

Hi, I'm setting up a Trino in production that will work together with DBT, Snowflake and SQL Server. I'm having problems with the "create table as select" command (used by DBT depending on the type of materialization).

The problem is: in Snowflake, the create for a table (+-16MM) using a warehouse (X-Small) runs in 10s, the same table using the same warehouse in Trino takes several hours. The planned use of Trino is to extract data from SQLServer and Snowflake and write it to Snow, through DBT commands (hence the tests with create table).

I would like to know if I can optimize using Trino's settings for this process. Regarding the cluster size, I've seen that Trino is very small [4 workers (CPU: 2 Memory: 9GB), 1 coordinator (CPU: 1 Memory 4GB)], and I'm already thinking about resizing the cluster to increase the pods. So far, the plan is: [8 workers (CPU: 8 Memory: 16GB), 1 coordinator (CPU: 4 Memory 8GB)], but I believe that this optimization alone, despite being necessary, is not enough. Could you shed some light?