Credible service selection in cloud environments

With the development of the Internet, cloud computing has become the most popular paradigm for on-demand provision of computing resources and service solutions. Due to the high flexibility of cloud computing, a myriad of cloud-based services are designed and implemented to satisfy consumers’ diverse needs, and thus new challenges have arisen in cloud service selection. One big challenge is how to select the most suitable cloud service for potential consumers according to their customized requirements. Another big challenge is how to perform cloud service selection with high effectiveness and accuracy, i.e., identifying unreasonable assessments and eliciting credible assessments. This thesis aims to systematically investigate the above two challenges of cloud service selection. The main contributions of this thesis are summarized as follows: In prior studies, cloud service selection usually depends on quantitative performance analysis without considering cloud consumers’ opinions on service performance. This causes a problem that some vital performance aspects, which can hardly be evaluated through objective monitoring and testing, e.g., data privacy and after-sales services, are ignored in cloud service selection. To solve this problem, we propose a novel model of cloud service selection by aggregating both subjective assessments from ordinary cloud consumers and objective assessments extracted through quantitative analysis. By comparing and aggregating such assessments, the result of service selection can reflect the overall quality of a cloud service with less bias caused by unreasonable assessments. We further consider the contexts of cloud assessments and cloud service requesters in our proposed model. In this model, a cloud consumer is allowed to specify under what condition (e.g., specify a particular location or a particular period of time) he/she would like to consume a cloud service. Then the service selection is carried out based on the consumer’s context. In this way, our cloud service selection model can more effectively reflect potential cloud consumers’customized requirements. In order to improve the accuracy of cloud service selection, we propose a novel approach to evaluate the credibility of cloud assessments. Considering the dynamic nature of cloud services, the proposed approach is based on the continual assessments over time, and thus has the ability to not only evaluate the dynamic performance of cloud services, but also resist user collusion of providing malicious assessments. Through this approach, more credible assessments are selected as input for further cloud service selection. In addition to the assessment credibility evaluation, we have found another way to further improve the accuracy of cloud service selection. We propose a novel incentive mechanism, through which a cloud user would have sufficient incentives to provide continual and truthful assessments of cloud services. This would benefit both the dynamic evaluation of cloud performance and the further cloud service selection. Furthermore, the proposed incentive mechanism allows users to provide uncertain assessments when they are not sure about the real performance of cloud services, rather than providing untruthful or arbitrary assessments which may greatly affect the accuracy of service selection. All the models, approaches and mechanism proposed in this thesis have been validated and evaluated through sufficient experiments and theoretical analysis. The results have demonstrated that the proposed approaches and mechanisms outperform the existing work of cloud service selection.