Appropriate mechanisms for the support of optional processing in hard real-time systems

Abstract

It has been recognised that future hard real-time systems need to be more flexible than current scheduling theory permits. One method of increasing flexibility is the incorporation, at run-time, of optional components into processes with hard deadlines. Such components are not guaranteed offline, but may be guaranteed at run-time if sufficient resources are available. This is achieved by providing mechanisms within the kernel for run-time monitoring of spare processor capacity and its subsequent assignment to requesting processes. This paper examines these mechanisms within the context of fixed priority preemptive scheduling.<>