Feedback Control Real-Time Scheduling: Framework, Modeling, and Algorithms

Abstract

This paper presents a feedback control real-time scheduling (FCS) framework for adaptive real-time systems. An advantage of the FCS framework is its use of feedback control theory (rather than ad hoc solutions) as a scientific underpinning. We apply a control theory based design methodology to systematically design FCS algorithms to satisfy their transient and steady state performance specifications. In particular, we establish a dynamic model and performance analysis of several feedback control scheduling algorithms, which is a major challenge and key step for the control design of adaptive real-time systems. We also generalize a FCS architecture that allows plug-ins of different real-time scheduling policies and QoS optimization algorithms. Based on our model, we identify different types of real-time applications where each FCS algorithm can be applied. Performance evaluation results demonstrate that our analytically tuned FCS algorithms provide robust steady and transient state performance guarantees for periodic and aperiodic tasks even when the task execution time varied considerably from the estimation.