The implementation of hierarchical schedulers in the RED-Linux scheduling framework

Hierarchical schedulers are useful to integrate different scheduling paradigms together. The original RED-Linux general scheduling framework does not support hierarchical schedulers efficiently because the dispatcher cannot tell whether a job is an aperiodic job or a real-time job. In the work reported in this paper, we add an extra parameter, the group number, to the RED-Linux scheduling framework in order to identify the type of jobs. This mechanism does not introduce any overhead to normal real-time tasks and only a constant overhead per job for hierarchical jobs. We discuss how to implement hierarchical schedulers and how to use this extension to support sporadic schedulers. We also discuss various versions of the sporadic server algorithm.