Contents 1 Introduction
Recent years have seen the rapid growth of Distributed Real-time Embedded (DRE) applications executing in unpredictable environments in which workloads are unknown and vary significantly at runtime. Such systems include data-driven and open systems whose execution is heavily influenced by volatile environments. For example, task execution times in vision-based feedback control systems depend on the content of live camera images of changing environments [13]. Likewise, supervisory control and data acquisition (SCADA) systems for power grid control may experience a dramatic load increase during a cascade power failure [9]. Furthermore, as DRE systems become connected to the Internet, they are exposed to load disturbances due to variable user requests and even cyber attacks [9]. As such systems become increasingly important to our society, a new paradigm of real-time computing based on Adaptive Quality-of-service (QoS) Control (AQC) has received significant attention. In contrast to traditional approaches to real-time systems that rely on accurate knowledge about the system workload, AQC can provide robust QoS guarantees in unpredictable environments by adapting to workload variations based on dynamic feedback. A key advantage of AQC is that it adopts a control-theoretic framework for systematically developing adaptation strategies. This rigorous design methodology is in sharp contrast to heuristic-based adaptive solutions that rely on extensive empirical evaluation and manual tuning.
