New platforms for parallel and distributed computing, such grids, are emergent structure build by collecting resources provided by volunteers individuals or organisations. These resources are not always available. They are in fact subject to disturbances related to uncertainty on start and end availability time. In order to design schedules adapted to these platforms we aim to optimise, in addition to the classic objective function, a new criteria that characterise the ability of the schedule to absorb the effect of the perturbation (stability). In this work, we study the problem of scheduling under availability resource constraints and uncertainty on the events occurrence dates. We initially study the elementary case where only one unavailability is allowed per machine and where the duration of the availability is uncertain. We then generalize to the general case when multiple unavailabilities are allowed by machine and where uncertainties are related on the events occurrence dates. For all these problems, we design schedules based on the slack technique, and that optimise both performance and stability. A wide simulation campaign of the designed heuristics lead to the identification of configurations that produces satisfactory results.
