Yes, that is usually the goal of performance testing. In general there are two sub-cases:
-
People have an idea of the required performance of a system, and want to gain some idea of which system configuration (number of routers, MPs, etc) will meet the performance objective with an appropriate margin of safety. For example, I might have a good estimate that the total API load driven by clients will be 1750 TPS at peak. Suppose I can measure that the max performance of the 9-node system is 2000 tps, with my given max TP99 latency. I might decide that I need more nodes in the system in order to allow a larger margin of safety, to allow load spikes or the ability to “lose” a node (for example for maintenance) without incurring a service disruption. So based on this I might configure the system to use 11 nodes, rather than 9. That system might give 3000 tps, but it’s the right margin of safety for me.
OR
-
They have a system and want to estimate the peak performance of the system, so that they can plan for future upgrades should API volume and load increase at some later point in time. For example, I have a 9-node system and a new team within my company wants to host their APIs and this will bring 120TPS into the system. Do I have confidence that the 9-node system with the existing API volume, can also accommodate the new volume? With a clear idea of the capacity of the existing system, you can make that decision confidently.