+1, the best metrics we tracked here were actually “events consumed per second per core” and “consuming segments per core”. you have 3 variables here: partitions, num servers, and num cores per server; so it’s easiest to pick numbers if you can constrain one of them