Research On Trust Evaluation Mechanism For Service-Oriented Grid

With the developing of Grid computing, Grid has extended its territory from traditional computing Grid to service-oriented Grid, aiming to realize e-Science and e-Business at the global level. Users pay more and more attention to QoS demand. Community-based adaptive metrics such as trust and reputation serve as building blocks to support QoS requirements. There is an urgent need for a new trust mechanism for service-oriented Gird with scalability, dynamic adaptability and robustness in a whole.First, an extensible QoS information management framework based on service-oriented Grid architecture is proposed in this thesis. It emphasizes that QoS model is customizable and the technologies on collecting, storing, measuring QoS information are introduced.Current research on QoS trustworthiness evaluation lacks of considering application's QoS constraints and differentiated service. To ensure this, a Bayesian Network based approach is presented to evaluate the trustworthiness of Grid service. The approach employs performance history to construct an evaluation model, which ranks the services in terms of their trustworthiness of QoS information and considers degree of consistency in QoS provision. Simulative results show that the proposed approach can effectively enhance trustworthiness of service information and improve reliability of service selection.A scalable trust model named QoSTrust is proposed based on current trust and reputation research in Grid and P2P network. In order to reduce the subjectivity of user's rating, two new concepts, namely Satisfactoriness of QoS and Trustworthiness of Service Rating, are defined in the proposed model, which are used to evaluate delivered QoS that fulfills agreed QoS in SLA quantitatively and trustworthiness of users' rating respectively. In trust inference, direct trust is defined with QoS constraints to meet user's QoS requirements better and VO level' trust relations are introduced to reduce the overhead of the reputation iterative algorithm. Simulative results show that the establishment of trust is more efficient. Also, when using the proposed model for service selection, it outperforms other models in terms of query success rate and user's satisfaction.