A business critical application in my portfolio had an on-premise PostgreSQL database that was long overdue for a version upgrade. After doing discovery for the project, we found out that DB licensing and new VMWare infrastructure costs would add up to a significant amount for this database cluster.
At that point we did a cost analysis to move the on-premise PostgreSQL cluster to AWS RDS. The cost of RDS (+ the required security and compliance assessments) came at ~50% less than the on-premise option. I decided to proceed with the RDS option.
Although we initially observed some latency between the on-premise application and the cloud RDS database, we were able to address most of these by refactoring the application's database invocation logic (mostly by optimizing the Hibernate layer).