Efficient EDF Implementation for Small Embedded Systems

Modern embedded systems are often required to execute under stringent real-time constraints to satisfy high performance requirements. When the available computational resources are scarce, an efficient implementation of the scheduler is of crucial importance for containing system costs. In this paper, we present an efficient implementation of the Earliest Deadline First (EDF) scheduler, which is typically considered difficult and expensive to implement with respect to fixed priority algorithms, such as the Rate Monotonic (RM) scheduler. The higher resource utilization and the greater flexibility in handling aperiodic requests and overload conditions make EDF highly desirable for real-time embedded systems. We show that by using suitable kernel mechanisms for time representation, EDF can be effectively used even in small microprocessors for achieving precise time representation and increasing system utilization. Another important issue for adopting new schedulers in the market is their compliance with existing standards. For this reason, in this paper we discuss how to achieve a compliance with the OSEK/VDX API, widely used on automotive embedded systems, and we show how commercial kernels can integrate an EDF scheduler while maintaining an OSEK compliant API. Finally, we provide some performance measurements to show that implementing and EDF scheduler on a commercial OSEK-like kernel, such as ERIKA Enterprise, only introduces negligible overhead.