Scheduling analysis of tasks constrained by TDMA: Application to software radio protocols

The work, presented in this article, aims at performing scheduling analysis of Time Division Multiple Access (TDMA) based communication systems. Products called software radio protocols, developed at Thales Communications & Security, are used as case-studies. Such systems are real-time embedded systems. They are implemented by tasks that are statically allocated on multiple processors. A task may have an execution time, a deadline and a release time that depend on TDMA configuration. The tasks also have dependencies through precedence and shared resources. TDMA-based software radio protocols have architecture characteristics that are not handled by scheduling analysis methods of the literature. A consequence is that existing methods give either optimistic or pessimistic analysis results. We propose a task model called Dependent General Multiframe (DGMF) to capture the specificities of such a system for scheduling analysis. The DGMF task model describes, in particular, the different jobs of a task, and task dependencies. To analyze DGMF tasks, we propose their transformation to real-time transactions and we also propose a new schedulability test for such transactions. The general analysis method is implemented in a tool that can be used by software engineers and architects. Experimental results show that our propositions give less pessimistic schedulability results, compared to fundamental methods. The results are less pessimistic for both randomly generated systems and real case-studies from Thales. These results are important for engineering work, in order to limit the over-dimensioning of resources. Through tooling, we also automated the analysis. These are advantages for engineers in the more and more competitive market of software radios.