Recently I wrote a test harness that had a “discovery” mode. That is, given a target service level agreement (SLA), what transaction pacing or throughput did the load harness have to apply in order to achieve the SLA response times. Now my maths is a tad scratchy, but I knew that I could somehow figure out the equation to determine how much the load balancer should adjust throughput in order to achieve the desired response time. In my mind I pictured a horizontal line that is the target, and a sinusoidal line that is the actual response time that varies according to load. If I could just figure out how to dampen that curve, I would have the magic [...]