Toward a deeper use of feedback control in the design of critical computing system components

Feedback controllers are typically applied to com- puting systems by acting on some quantities (“tunable parame- ters” in the computer science lexicon, think e.g. of a packet drop rate) in order to attain some goal (e.g., a required bandwidth allocation). In other words, control loops are closed around the computing system, that exposes tunable parameters as control actuators, and allows to measure some performance metrics. In the authors’ opinion, the “classical” use of control methods sketched above heavily limits the achievable results. Indeed, many critical parts of computing systems should not be “con- trolled”, but rather “re-designed in the form of controllers”— quite a novel approach and an interesting research area. This manuscript formalises the statement above and illustrates some results, including a couple of application examples, to demonstrate the advantages of the proposed design approach.