We have Drill on YARN configured on our MapR cluster and we are using Drill's JDBC interface to fetch data from a DB Client.
To improve query performance I am using Connection Pooling, so that a new connection is not created for each query request.
If we have 1000 such DB clients, each maintaining a connection pool of 20 connections each, can a single Drill on YARN handle around 20,000 simultaneously open JDBC connections?
If yes, how can I do performance tuning considering the above deployment? Will "planner.width.max_per_node" help?