A novel framework for design and analysis of embedded software with variable constraints

点击翻译按钮获取中文摘要

Modern embedded systems need design considerations for applications that may require changing their behavior over time. To run an application in different situations, embedded systems need to consider different user-driven constraints such as safety, security, or usability, and system-oriented constraints such as cost, timing, energy consumption, and throughput. However, these constraints may oppose each other, and therefore improving one means accepting deterioration in another. For example, physical redundancy might increase safety while also increasing the attack surface and thereby reducing security. Balancing the constraints provides a good performance of the system; however this might not work well for applications in situations that require optimization of a particular constraint or a combination of constraints. The requirement to operate an embedded system in different configurations associated with variable constraints fosters the development of a novel framework composed of a set of innovative design and analysis methods and tools. Several research problems exist in designing an embedded software with variable constraints. First, selecting an efficient set of design elements and their configurations is challenging because of the evolving requirements of tasks and messages in an application and the presence of a wide variety of embedded systems. Secondly, proper analysis on tolerable delays for variable constraints is required to reduce overprovisioning of resources. The framework will map workloads into different configurations or modes, where each configuration optimizes a particular or a combination of constraints. However, mapping variable constraints to multiple modes may require adaptation of an existing scheduling strategy or a new one. Moreover, an application can provide acceleration of performance upon multiprocessor platforms, but the mapping of workloads to processors requires rigorous analysis. Finally, analytics is necessary to identify uncertainties and generate recommendations for configuration changes or adaptations. The proposed framework for design and analysis of embedded software comprises the following research activities: (1) efficient selection of design elements and their configurations, (2) analyzing delays for workloads with variable constraints, (3) mapping variable constraints to multiple modes, and (4) processing of events and analytics. Throughout the research process, students will receive specialized training in high-demand embedded software and gain skills which will introduce new embedded software design practices, achieving a closer coupling of user- and system-oriented constraints. The outcomes promise to improve Canadians' ability to gain insights from economically and socially important case studies in domains from consumer electronics to the Internet of Things.