I have been working on a customer database benchmark where latency and throughput are critical. The customer provided me with some sample code in Python which used the cx_Oracle driver. I was concerned that either the Python language or the cx_Oracle SQL driver would be the bottleneck and whether I would need to re-write the benchmark in 'C' using OCI or ODBC.
My target was 300K queries per second, where each query consisted of a four table join + 2 single table queries.
After some SQL tuning [ie using the TimesTen Index Advisor] I was able to achieve 11.3 million queries per second on a single VM in Oracle Cloud using 27 database connections with Oracle TimesTen 220.127.116.11.0.
Given that 11.3 million is significantly faster that 300K, I concluded that using Python with the cx_Oracle driver was 'fast enough'!
As a bonus, the exact same Python source code worked for the Oracle Database and Oracle TimesTen. ie I just needed to change my tnsnames.ora service name to point to either the Oracle or TimesTen Database.
Disclaimer: These are my personal thoughts and do not represent Oracle's official viewpoint in any way, shape, or form.