When you start benchmarking, you will probably encounter some issues, such as:
- A network setup not correctly done
- A bug in a framework/a library/your application/the injector
- A remote service or database not absorbing enough load, and so on
In all potential cases you may encounter—given any brick of your software can have an issue—you should be able to have somebody you can call to help you fix the issue, or at least evaluate it quickly. This is particularly crucial if part of the benchmark costs some money (if you are renting some machines, consulting, and so on). The idea here is to be able to get rid of any blocker as fast as possible to not waste time on details.