Multiprocessor Task Scheduling with Probabilistic Task Duration

Article concentrate on so-called multiprocessor task scheduling problem with uncertain (random) time of tasks duration. Multiprocessor scheduling can be perceived as a tool for improving dependability of the system by hardware and software redundancy. Our overall aim is develop more precise description of various configuration of embedded systems). Our aim is to evaluate the model of task duration distribution on the result of scheduling. We compare two models of the task: normal distribution and Erlang distribution. The latter model is considered as more suitable for multiprocessor scheduling, which reflects more accurately reality. We use MVA (Mean Value Analysis) methodology in the research with the application of modified Muntz-Coffman algorithm. The article is an extension of previous research of the author, considering uncertain task duration in different models. Computational experiments compared results obtained for both distributions in this stochastic model.