Benchmark analysis of a control-theoretical approach to feedback scheduling

A feedback-based approach to process scheduling in a single-processor system was recently proposed, that deeply relies on the control theory for its entire design. In this work a literature benchmark is used to compare said approach to classical ones, with particular reference to the real-time case. Two are the main results. First, in the case of a schedulable process set, the proposed “control-theoretical” design yields similar performance with comparable or lower time complexity with respect to classical approaches. Second, and more relevant, if the required CPU utilisation transiently exceeds the unity, the proposed approach allows for a significantly better system recovery (measured in terms of deadline misses) once said utilisation returns feasible.