Energy-efficient Computation And Modeling Advancements In Multi-processor Scheduling For Real-time Embedded Systems

Advancements in design and objectives for real-time embedded systems have created some challenges.These challenges mainly focus on maintaining and improving the quality of services,improving the system's response time,and decreasing power consumption.Moreover,portable systems that heavily rely on batteries for energy are required to limit the battery size,and demand in the application increases over time.On the other hand,it is necessary to know the worst-case response time of the system before the actual execution,specifically for safetycritical applications.Worst-case response time is required to perform ahead of the practical implementation because real-time systems can fail to map realistic models without missing any deadline—moreover,the cost of fixing the potential error increases later in the design stages.This dissertation proposes two algorithms on the operating system level to reduce the power consumption in multiprocessor systems to support our thesis.These techniques are called MODE and Ra SAM based on dynamic power management and dynamic voltage and frequency scaling.These techniques use the Global EDF scheduling algorithm as the backbone and reduce power consumption.As a scope of this dissertation,these techniques are implemented on simulation tools for real-time systems.On the other hand,a novel approach for measuring the worst-case response time of periodically triggered task-chains in automotive embedded systems is proposed.This technique is based on performing compositional performance analysis in distributed systems.The study presents a holistic approach and gives a general overview of the system's worst-case response time when multiple ECUs are connected over the CAN bus.The analysis is demonstrated through simulation of task-chains by using compositional performance analysis framework in Python.