Tuning your webservers, without any metrics, can be tricky. Often neglected is the fact that a web application runs on top of a large stack, starting with hardware powering all the software pieces on top. And all those different layers come with their own configuration. So what is the best way to set this all up, maximizing performance?
First-class stress testing
We simulate a realistic workload with a configurable number of virtual users. To see how the (virtual) hardware reacts under stress, we can monitor a large range of counters including CPU, memory, I/O and power usage with our monitoring toolset. This way we are able to determine how the application performs, and bottlenecks can be pinpointed.