Today, I gave a tutorial during a spring school on the energy efficiency of distributed systems targetting doctoral students. Here is the summary that I gave for that presentation:

This talk aims at answering the nerve-racking question "how to conduct good experiments leading to a good science" in the domain of large-scale distributed systems such as HPC or Cloud systems.

We shall see that it is possible to build real scientific instruments based on the simulation of mathematical analytic models. Addressing the resulting challenges turns out to be an exciting scientific adventure, leading to unexpected benefits: the methodological improvements that simulation mandates reveal beneficial regardless of the used scientific instruments.

Computer science then become a regular computational science, taking advantage of any available tool to address its challenges.