Philipp K. Janert - Feedback Control for Computer Systems

Feedback works by constantly comparing the actual behavior of a system to its desired behavior. If the actual behavior differs from the desired one, a corrective action is applied to counteract the deviation and drive the system back to its target. This process is repeated constantly, as long as the system is running.

One appealing feature of feedback control is that it requires relatively little knowledge about the controlled system. As long as one knows which direction to nudge the system when it has gotten off course, one can build a feedback loop. For this reason, feedback is an attractive technique for controlling large, complex, and opaque systems.

Moreover, feedback systems are self-correcting even in the presence of external disturbances. Because the systems behavior is monitored and adjusted all the time, a feedback system naturally and automatically responds to changes in operating conditions. No need to make special provisions to activate additional servers for rush hour: the feedback controller will notice the increase in load and spin up further instances until the desired quality of service is met. It will also take the instances down again, once the rush has passed, in proportion to the lightening load.

One needs to be careful, though, since control actions that are either too large or improperly timed can overcorrect a disturbance. Instead of reducing the difference between the desired and the actual behavior, such control actions replace a deviation in one direction with another deviation in the opposite direction. In the worst case, the amplitude of these deviations grows with each step until the system becomes dysfunctional or, often literally, blows up!

To avoid this outcome, control theory has developed specific experiments for analyzing a systems behavior. The results from these experiments can then be used to design and tune control loops that are safe to operate and that will track a reference value accurately and reliably.