writes "I have recently performed load testing for a client's web application and have been pondering the results and how to relate them to real world usage. The application is simple — login, enter details on a form and submit the form. It is the actual processing done on submitting the form that is interesting and has previously had performance issues. I can simulate a peak usage of 100s of browsers where the form is continuously submitted but how can I extrapolate the results of this test to thousands of users who will all on a single day submit the form once at some time through the day? What techniques statistical or otherwise do people use to determine the likelihood of the tested peak load being exceeded?"