An Efficient Algorithm for Scheduling Jobs in Volunteer Computing Platforms

We propose in this paper a new algorithm for scheduling independent jobs in volunteer computing platforms. In such platforms, the resources are not continuously available over time. Moreover, the beginning and finishing times of the resource availability periods are subject to some uncertainties since the machines are directly administrated by the owners (and thus, there is no global centralized control). The performance of the applications is achieved by a suitable adaptation of the algorithms to the volunteer environments. We propose an efficient method that optimizes the performance and reduces the impact of disturbances by means of stability. Stability gives the guarantee that the disturbed solution does not differ too much from the initial solution. It is obtained on the basis of a reputation mechanism of the resources that makes the scheduling decisions more adequate. The quality of our algorithm is demonstrated by a campaign of experiments using simulations of actual traces of BOINC where the results are compared to other existing reference algorithms.