We have it easy, but do we have it right?

We show two severe problems with the state of the art in empirical computer system performance evaluation, ob-server effect and measurement context bias, and we outline the path toward a solution. ©2008 IEEE.