Distributing Criticality Across Ada Partitions

Abstract

A number of real-time systems combine strict (hard) and non strict (soft) timing requirements. In such systems, the hard timing requirements must be guaranteed, whereas tasks with soft requirements may miss some deadlines with no harmful consequences. One possibility to organise such systems is to run the critical tasks in the kernel space, at the highest priority available, while executing the soft tasks in user space, where they will use the remaining execution time after servicing the hard tasks. A recently produced porting of Ada to Real-Time Linux makes this architecture feasible.

In this paper we propose a method for splitting a hard/soft real-time application into criticality-based partitions: the hard partition will run on RTLinux and the soft partition (or partitions) will be placed in the user space of Linux. The use of the Distributed System Annex enables the programmer to automatically place each partition in the right space and allows to hide the communication details between partitions. We also outline a strategy to reduce the average response time of soft tasks.

This work has been partly funded by the European OCERA project [1] and by the Spanish Comisión Interministerial de Ciencia y Tecnologí a project number TIC-99-1043-C03-02.