Reviewing RDS cost break down

Some of the cost components depend on your configuration. In addition to CloudWatch metrics and Performance Insights that the bot mentioned as separately charged optional features, you might also set RDS to send logs to CloudWatch logs, where both the log vending and storage would incur some costs.

Since you're looking at costs and are running without commitments on on-demand capacity, you might also want to take the opportunity to grab some quick wins. db.m5.large is based on the Intel x86-64 processor architecture and fifth-generation EC2 instances launched in November 2017. db.m6i.large and db.m6i.xlarge are sixth-generation equivalents also based on Intel's processors.

AWS has their own Graviton line of processors optimised for cloud workloads and based on the Arm64 processor architecture. With RDS, switching to them is super easy. You can switch between Intel and Graviton processors just as you could between two Intel processors: simply change the "instance type" selection for your database instance. RDS runs exactly the same software on both platforms, so for your applications, there's no difference between the processor options. It's even possible to switch an RDS cluster from Intel to Graviton one node at a time; since the software is the same, one cluster node can seamlessly run on Intel while the other runs on Graviton.

AWS's general guidance is that the price-performance ratio can be up to 40% better over the comparable instance generation using Intel's processors, and given that your instance types are older ones, the improvement may even be greater. The m6g/c6g/r6g instance families use Graviton2 processors, and m7g/c7g/r7g use newer, more powerful Graviton3 processors. Graviton4 doesn't seem to be launched for RDS yet in us-east-2, but when it becomes available, it'll have the 8g suffix in the instance family identifier and again more power per dollar.

Without considering the possibility that with the more powerful processors, you might be able to switch down to smaller instance families/types (such as from m5 to t4g or from m6i.xlarge to r7g.large), simply mapping your current Intel instances to the correspondingly sized Graviton instance types, you'd get cost savings right away. Quoting list pricing at https://aws.amazon.com/rds/postgresql/pricing/:

db.m5.large and db.m6i.large (both at $0.356/h) would map to:

• db.m7g.large: $0.337/h (Graviton3)
• db.m6g.large: $0.318/h (Graviton2)

db.m6i.xlarge ($0.712/h) would map to:

• db.m7g.xlarge: $0.674/h (Graviton3)
• db.m6g.xlarge: $0.636/h (Graviton2)

Note that while the Graviton2 generation instances have cheaper price points, they are less powerful than Graviton3, so considering right-sizing, Graviton3 may easily outperform them cost-wise.