Kernel Overheads Research Articles

Precise response time analysis for Ravenscar kernels

The Ravenscar Profile defines a subset of the Ada95 tasking constructs which can be implemented using a small, reliable kernel. One of the benefits of this approach is to improve the timing analysis because non-deterministic and non-analyzable features are excluded. But to perform a precise schedulability analysis of a Ravenscar compliant application, the kernel overheads must be taken into account. Therefore, a set of metrics and documentation requirements has to be defined in order to model the kernel influence in response time analysis. In this paper the main components of the response time of periodic and sporadic tasks running on a Ravenscar kernel are identified, and the set of kernel metrics which are required to perform a precise timing analysis is identified.

Keynote paper: A performance analysis of a hard real-time system

This paper analyses the performance of an Attitude and Orbital Control System for the Olympus Satellite. The system was designed using HRT-HOOD and implemented in Ada 83 augmented with some of the facilities of Ada 9X. The program consisted of multiple tasks communicating via protected shared data areas. A worst-case execution time analysis tool was used to estimate the execution time of each task, and Deadline Monotonic schedulability analysis was used to predict the system performance, taking into account all kernel overheads. An instrumented version was also produced to gather statistics on the actual run-time performance.