What tools or best practices exist to demonstrate a new server instance is "as capable" as an older one?
We're moving from an EC2 hosted instance of MS SQL Server to an RDS instance, and have been tasked with "proving" that the new platform won't cause any regressions.
The naive idea is to somehow replicate a day's worth of load, and record the elapsed time for all the queries. Then, if the worst 0.1% of the queries run no slower, we're good.
Is that logic good, or is there a flaw, or is there a better/standard way?
What would I need to do on the old server to "capture" a day's worth of load? (Queries, parameter values, session id so that I can simulate the correct number of concurrent connections, login name, elapsed query time, etc).
What else am I missing?
(Another niave idea was to put a "load balancer" on to "replicate" the queries to both instances, and log the query times. Is that even possible?)