Perspectives on quality of service in distributed and embedded real-time systems

Abstract

As a consequence of technological advancements, a trend towards the development of smart cities has emerged, i.e., towards urban areas that comprise a multitude of sensors, actuators as well as computation and communication resources. Being integrated into buildings, infrastructure elements, and other objects, these components constitute a large and heterogeneous distributed hardware platform. Traffic participants and other actors of a smart city can use this platform in an on-demand fashion to make use of advanced functionalities such as, for instance, smart means of transportation. In fact, vehicles of different levels of autonomy rely on a smart city's distributed infrastructure when performing sophisticated operations that come with specific quality of service (QoS) requirements, including a multitude of parameters such as timing and reliability constraints. Against the background of a shared, heterogeneous hardware infrastructure, however, guaranteeing the satisfaction of QoS requirements and, thus, ensuring the operations' correctness is an intricate matter.

This dissertation addresses selected challenges arising in the context of smart cities, focusing on the underlying distributed system as well as on individual systems interacting with it. All challenges contemplated are related to the notion of quality of service and aim to either guarantee the satisfaction of applications' QoS requirements or to enable the system(s) to enhance the level of service provided to (specific types of) applications. Concretely, a concept of QoS contracts concluded between the distributed system and each executed application is proposed that allows to provide QoS guarantees and, moreover, to detect contract violations. An extension of this concept including applications with robustness requirements is provided as well. For individual systems, focusing especially on smart vehicles, recovery protocols are proposed that enable the system to safely offload parts of critical applications to a smart city's distributed system, even under unreliable connections, while ensuring the temporal correctness. In addition, an approach for the optimization of hardware message filters in controller area network is proposed by means of which the overhead due to unnecessary message inspection can be reduced, allowing to spend the saved resource capacity on the execution of other applications. All concepts and approaches contributed in this dissertation have been evaluated and shown to be effective.