Rodney Bone is a performance consultant that works for Revolution IT in Brisbane, Australia. He has graciously invested his time to share insights about performance testing with us. Please follow him on Twitter (@rod_bone) and tweet your thanks for his interview.

That's Rod in the picture at the right when he did some reserve time during the Brisbane floods over Christmas. The slow sign is just ironic.

What is your technical background?

Started as a software engineer with Accenture, where I was involved in the entire end to end SDLC including experience in the BA space.

Do you consider yourself more of a software developer or QA professional?

Now, definitely QA, I think todays generic developer is only one part of a large picture and Accenture’s model of exposing developers to the whole SDLA is one that should be encouraged by all companies.

How do you determine the load to apply to the target app during load testing?  

Ask the business, and ask as many people as applicable. Users, and BA’s have different opinions, so talk to them all. Once you nail the processes and the amount of transactions per hour run with that. Back that up with Log info as historical information is the best source of information. With new applications this is not always available.

Do you prefer using requests per second (RPS), concurrent users, or some other metric to define load?

They answer different questions. You can sometimes hit a server with the right RPS with one virtual user but that doesn’t tell you that the web server is up to serving the required amount of concurrent connections. I have tested an app where they plugged the Apache Load Balancer in out of the box and it only served 20 concurrent connections. FAIL! And you wouldn’t pick this up without concurrent users.