Analysis and redesign of the ‘TTC’ and ‘TTH’ schedulers

The Time-Triggered Cooperative (TTC) and TT Hybrid (TTH) schedulers have previously been described in the literature as highly predictable static schedulers with very low overheads, which are suitable for use in resource-constrained embedded software applications. Although several previous papers have considered these schedulers to a certain degree of depth, to date there has not been (i) a serious comparative investigation of the scheduler properties or (ii) the attempt to apply mainstream scheduling theory within their frameworks. As designs based around the TTC framework seem to be increasing in popularity, with operating systems based around these schedulers now commercially available, these points should be investigated; this is the focus of the current paper. The investigation reveals that although it can be said that the schedulers are predictable, their underlying operation seems to be dynamic – in the regular sense of task priorities – and employs the Earliest Activation First (EAF) priority assignment rule. In addition, with their present design the TTC/TTH scheduling overheads do not scale well, and can be excessive in comparison to alternative techniques. Motivated by these observations, the paper goes onto describe: (i) a re-design of the schedulers which allows the overheads to scale at an acceptable level, and improves the flexibility of the scheduler, (ii) the formulation and proof of an optimal tie-breaking rule for the schedulers, and (iii) the application of mainstream scheduling theory to yield a schedulability test of improved efficiency. The effectiveness of these modifications is illustrated with a small comparative study. Finally, the paper provides an overview of TTC and TTH properties in comparison to other popular scheduler architectures. These latter points should be of interest to practitioners in the field.