The global EDF scheduling of systems of conditional sporadic DAG tasks

The sporadic DAG task model exposes parallelism that may exist within individual tasks to the run-time scheduling mechanism, and is therefore considered a particularly suitable model for representing recurrent real-time tasks that are to be implemented upon multiprocessor platforms. This paper proposes and evaluates an extension to the model to allow for the concurrent modeling of conditional execution of pieces of an individual task, along with the modeling of intra-task parallelism. The Global Earliest Deadline First (GEDF) scheduling of systems represented in this generalized model is studied, and a GEDF-schedulability test is derived. With regards to GEDF scheduling it is shown that there is no penalty, in terms of worse speedup factor, in generalizing the sporadic DAG tasks model in this manner.