In order to avoid discovering our system at large scale the first time we'll need it, perform load testing at moderately large scale to assess:
- if scaling works properly
- end-to-end througput (with temporal overhead, preprocessing, machine provisioning and so on): "how many hours does it take to translate X million characters from language L to English"
- cost
In order to avoid discovering our system at large scale the first time we'll need it, perform load testing at moderately large scale to assess: