Work stealing based volunteer computing coordination in P2P environments

This paper aims at the evaluation of work stealing based Volunteer Computing (VC) coordination with the goal of confirming the scalability of VC for Peer-to-Peer (P2P) environments. Our previous work has successfully modelled work stealing for VC coordination and been evaluated by a few applications and a small number of machines. However, this paper argues that the evaluation of scalability of VC by the statistical data of real world applications or through mathematical modelling is either limited by the number of volunteer machines or difficult to achieve because of the peer churn in P2P opportunistic environments. This paper proposes a simulation model for the same VC functions but performing the virtual work so that the statistical data can be quickly obtained for the dynamic behaviors of a large number of volunteers. The initial evaluation results have demonstrated that the work stealing based VC coordination scales up to 10K volunteers against varying churn rate, communication cost and stealing granularity. The confirmation of scalability ensures that VC can be effectively applied to P2P opportunistic environments.