Robust Scheduling of Real-Time Applications on Efficient Embedded Multicore Systems
Michael Deubzer · mediaTUM – the media and publications repository of the Technical University Munich (Technical University Munich) · 2011
The increasing demand of computing capacity in embedded real-time systems, especially in the automotive powertrain domain, has been satisfied so far by increased processor frequency of singlecore processors.Due to thermal dissipation, increasing processor frequency is technically limited.Multicore processors provide additional computing capacity at constant or even reduced processor frequency.However, most embedded real-time systems have tight real-time, efficiency, and robustness requirements, resulting respectively from time-critical interaction with other devices, high piece numbers, and realization of control functions.Multicore processor real-time scheduling approaches, namely partitioning heuristics for existing scheduling algorithms for singlecore processors and global scheduling algorithms for multicore processors, have to fulfill these requirements.This work discusses multicore processor real-time scheduling approaches as well as examination approaches of real-time, efficiency, and robustness properties for complex multitasking applications (task sets) on multicore processors in embedded real-time systems.Related work on multicore processor real-time scheduling mainly considers simplified task sets, i.e. periodic or sporadic task sets with constant task execution times and implicit deadlines.This work examines multicore processor real-time scheduling approaches for more complex practical task sets, common in automotive systems, with heterogeneous task activation patterns, variable task execution times, and hard or soft explicit task deadlines.The work considers both global scheduling algorithms and partitioning heuristics for local scheduling algorithms for symmetric multicore processors.