ABSTRACT
This chapter describes the design choices related to the implementation of a partitioned scheduler for allocating the computing resources to the different threads in the system. The number of operating system (OS) threads onto which a real-time task is mapped depends on mapping decisions. The chapter describes the overall schedulability analysis of systems executing within the P-SOCRATES framework. In a traditional partitioned scheduler, OS threads are statically assigned to cores, so that no thread may migrate from one core to another. Earliest deadline first (EDF) is known to be optimal for scheduling a collection of jobs upon a preemptive uniprocessor platform, in the sense that if a given collection of jobs can be scheduledto meet all deadlines, then the EDF-generated schedule for this collection of jobs will also meet all deadlines. In our scheduling framework, the global scheduler will therefore consist of a fixed-priority scheduling algorithm.
