Achieving Reliable And Timely Task Execution In Mobile Embedded Applications

摘要

We consider the real-time scheduling of the local processing tasks of a distributed sensor fusion application. The particular problem lies in the environment dependent and hence unpredictable execution times of these tasks, which do not allow using worst-case values. The TAFT (time-aware fault-tolerant) scheduling concept works with expected-case execution times, yet still achieves a predictable timing behavior. It aborts task instances that would miss their deadline and ensures a timely exception handling. Using measurements from a prototype of the application, we illustrate the TAFT concept and show how application-inherent redundancy can be exploited to tolerate the task abortions.