Performance between Dremio JDBC and ODBC driver

Do you have any benchmark about which driver can achieve better performance?

From my experience, it seems JDBC driver has better performance than pyodbc by Python.
For large query such as 2 millions records with reflection enabled, JDBC driver is 3x or 4x faster than ODBC.

We use dremio-jdbc-driver-3.1.3-201902130614340856-5cd4808.jar and odbc driver to compare.

Would it also be possible to not bump up the JDBC version with every new release if nothing in the driver has changed? Having a stable version when building java apps with Maven Ivy dependencies on the jdbc jar file is preferred.

There is currently no benchmark regarding ODBC vs JDBC. That said, I would expect comparable performance between both of them as they both use the same wire protocol, so a 4x speedup seems suspicious.

As for not bumping the JDBC driver version, I am afraid it is not quite feasible. The JDBC client code is part of our main code base, and even if the features of the driver doesn’t change, it benefits internally of some improvements of our main code base. That said, older versions of the driver would still work with newer versions of the server (except for new features introduced), so up to you to keep using the same driver version.