Differentially private and truthful auction-based resource procurement for budget-constrained DAG applications in clouds

With the rapid proliferation of emerging scientific computing, more and more cloud users (CUs) with large-scale applications are seeking efficient executing paradigms in clouds. Following this trend, a cloud market with multiple cloud service providers (CSPs) is emerging, which aims to provide efficient cross-cloud services to CUs through the cooperation among CSPs. In the market, CUs can procure different virtual machine resources and their combinations from one or more CSP(s) in a pay-as-you-go way. In such a procurement process, it is challenging to achieve privacy protection and truthfulness of CSPs while enabling flexible resource combinations. As such, we design a differentially private and truthful auction-based resource procurement mechanism (DTARP) in clouds. Specifically, we first propose a budget-aware resource profile determination algorithm (BRPD) to obtain a set of resource profiles, in order to bridge the difference between application tasks and resources under the budget constraint. Next, we design a differential privacy-based winning CSP selection algorithm to choose a set of winning CSPs, and the payment calculation algorithm to calculate the payment for each winning CSP. Strict theoretical analysis proves that the proposed DTARP achieves differential privacy, truthfulness, individual rationality, computational efficiency, and approximate social cost minimization while guaranteeing the efficient completion of application tasks under the budget. Extensive simulation results also demonstrate the effectiveness of the proposed DTARP.