Towards Multi-criteria Volunteer Cloud Service Selection

Abstract

Volunteer cloud computing (VCC) have recently been introduced to provide low-cost computational resources to support the demands of the next generation IoT applications. The vital process of VCC is to provide on demand resource provisioning and allocation in response to resource failures, behavior of volunteers (donors, users) and dynamically changing workloads. Most existing work addresses each of these factors (reliability, trust, and load) independently. Finding the most reliable machine (e.g., the lowest hardware failure rate) does not guarantee that the machine is trustworthy or not loaded, and vice versa. To address these problems, this research proposed a model to select volunteer node (VN) based on three criteria: the trustworthiness of the volunteer, the reliability of the node, and the resource load. We use three different models to estimate the three factors. We used exponential distribution reliability to estimate the reliability of VN and neural network to predict VN resource usages. In addition, we propose a new version of the beta function to estimate trustworthiness. Then we apply multiple regression to weigh each factor and decide which factor will be most effective for preventing task failure. Finally, a VN is selected based on multiple criteria decision analysis.