Businesses can rise or fall on the margins created by application performance. Conversions, revenue, and traffic all increase dramatically as websites get faster. Giant after internet giant has produced metrics reaffirming the importance of fractional seconds of response time. Performance is not an easy problem to solve, though. Intuitively applying optimizations and hoping for the best produces poor results. Inferring application characteristics from operating system metrics can mislead. Only application instrumentation can provide deep insight into application performance. Instrumenting apps is hard, though, and interpreting the results can be even harder. Profilers and system metrics are the first tools most engineers reach for, but neither is very good at understanding the performance implications of application behavior. Understanding the performance of production applications requires tools that measure the performance of production applications. This sort of instrumentation comes in many forms. Developers often write their own monitoring endpoints. Etsy's statsd couples well with graphite as a tool to instrument an application by hand. Distributed tracing tools offer a different way of looking at latency, by putting it in the context of a request/response cycle. Each of these tools has a role in uncovering some aspect of application performance.