Temporal protection in real-time operating systems

Abstract

Real-time systems manipulate data types with inherent timing constraints. Priority-based scheduling is a popular approach to build hard real-time systems, when the timing requirements, supported run-time configurations, and task sets are known a priori. Future real-time systems will need to support these hard real-time constraints but in addition (a) provide friendly user and programming interfaces with audio and video data types (b) be able to communicate with global networks and systems on demand, and (c) support critical command and control services despite potential risks introduced by such added flexibility and dynamics. We argue that temporal protection mechanisms can be as beneficial in these systems as virtual memory protection. The processor reservation mechanism that we have implemented in Real-Time Mach, for example, provides guaranteed timing behavior for critical activities.<>