SSCS-λ: A Cellular Automata-Based Scheduler with Stochastic Update Based on the Neighbourhood States

Cellular Automata (CA) are discrete dynamical systems composed of a transition rule and many cells that can assume a set of states, the rule updates a cell according to the states of cells in the proximity of that cell. The study of CA rules to decide the distribution of program tasks to system processors focus on training a CA rule over a specific instance and use this rule to solve several others examples. The best performing CA-based scheduler to this Task Scheduling problem uses a stochastic CA with a random cell update. Here we study different probabilistic distributions to be used in the model update, we regard and mix two components: the uniform distribution and a distribution that increases the probability of states that appeared more often in the neighbourhood of a cell. We investigate five models, in which we vary the influence of each component in the mix. The results endorse that the best scheduling result is found by variations where the two components importance is similar, these variations outperformed the state-of-art CA-based model.