Real-time operating systems: Problems and novel solutions

Abstract

This work presents some methodologies for enhancing predictability in real-time computing systems, where explicit timing constraints have to be enforced on application processes. In order to provide an off-line guarantee of the critical timing constraints, deterministic and analyzable algorithms are required in all kernel mechanisms, especially involving scheduling, inter-task communication, synchronization and interrupt handling. This paper illustrates some problems that may arise in real-time concurrent applications and some solutions that can be adopted in the kernel to overcome those problems. In particular, task scheduling algorithms and resource management policies will be considered in detail, as they have great influence on system behavior. Finally, a novel approach will be introduced for handling transient overloads and execution overruns in soft real-time systems working in dynamic environments. These techniques provide efficient support to real-time multimedia systems.