What if you had all the data you needed to measure system performance and scalability at any tier, discover performance and stability problems before they happen, and plan for capacity by modeling the system’s behavior at greater load than you currently have? Now it is as easy as running tcpdump and processing the result with a tool. In this talk you will learn a quick overview of mathematical scaling models, how the inputs can be derived from observing query start and completion times, and how the mk-tcp-model tool from Maatkit can transform tcpdump output into the inputs you will need for tools such as Aspersa’s usl tool for applying the Universal Scalability Model.