So,

Previously, we try to understand how many request that can be completed within the SLO.

100 ms (SLO)/ 35ms/request = ~ 3 request (within the SLO) (100ms = 3 request). If we convert this to second, we can get that 1.000 ms (1 sec) = 30 request.

From this information, we know that current capability of an instance (30 request/sec) is exceed the expectation (24 request/sec), hence additional instance is not required.

is that right?

Please reply, thank you