| Abstract: |
We consider a battery-less real-time embedded system equipped with an energy harvester. It scavenges energy from an environmental resource according to some stochastic patterns. The success
of jobs is threatened in the case of energy shortage, which might be due to lack of harvested energy, losses originated from the super-capacitor self-discharge, as well as power consumption of
executed tasks. The periodic real-time tasks of the system follow a dual-criticality model. In addition, each task has a minimum required success ratio that needs to be satisfied in steady state.
We analytically evaluate the behavior of such a system in terms of its energy-related success ratio for a given schedule. Based on these results, we propose a scheduling algorithm that satisfies
both temporal and success-ratio constraints of the jobs, while respecting task criticalities and corresponding system modes. The accuracy of the analytical method as well as its dependence on the
numerical computations and other model assumptions are extensively discussed through comparison with simulation results. Also, the efficacy of the proposed scheduling algorithm is studied through
comparison to some existing non-mixed- and mixed-criticality scheduling algorithms. |
