When analysing hard real-time systems for their performance properties, most scheduling models either ignore the timing characteristics of the underlying run-time support kernel entirely or have a very simple model of its behaviour. However, when engineering an actual implementation, it is necessary to tailor the scheduling theory so that the actual performance characteristics of the run-time system are adequately addressed. Failure to do this may result in a "guaranteed" task missing its deadline. In this paper we investigate an actual Ada run-time kernel and give performance measures for clock handling and delay queue manipulation. The impact these measures have on the scheduling models are discussed.
Download Full PDF Version (Non-Commercial Use)