goal (﴾or at least didn’t stay on track with the goal)﴿ • Not enough stress on the system • Too much noise in the measurements from the rest of the application to notice subtleties in the data • Skewed measurements by only looking at the case it was designed to work well for • Didn’t measure the full system
Create a separate benchmark app and measure in real life scenarios (﴾in the real app)﴿ • Put the system under heavy stress • Look at cases where the system would reasonably perform poorly • Profile the whole system