In the context of Grid computing, reputation-based trust management systems are playing an increasingly important role for supporting coordinated resource sharing and ensuring provision of quality of service. However, the existing Grid reputation-based trust management systems are considered limited as they are bounded to esoteric reputation-based trust models encompassing predefined metrics for calculating and selecting trusted computing resources and as a result, they prevent external involvement in the trust and reputation evaluation processes.
This thesis suggests an alternative approach for reputation modelling founded on its core argument proclaiming that reputation is a subjective matter as well as context dependent. Consequently, it offers a synergistic reputation-policy based trust model for Grid resource selection. This exoteric trust model introduces a novel paradigm for evaluating Grid resources, in which Grid client applications (e.g. monitoring toolkits and resource brokers) are endeavoured to carry out an active participation in the trust and reputation evaluation processes. This is achieved by augmenting the standard reputation queries with a set of reputation-policy assertions constituting as complete trust metrics supplied into the reputation algorithm. Consecutively, the Grid Reputation-Policy Trust management system (GREPTrust) provides a concrete implementation for the trust model and it’s underlying artifacts whilst the GREPTrust testbed provides an adequate infrastructure for comparing the reputation policy trust model with a production available esoteric model (GridPP).
Based on a computational finance case study, an internal workflow simulation utilises the GREPTrust testbed in order to empirically assess the criteria by which the synergistic reputation-policy based trust model outperforms esoteric trust models regarding resource selection and consequently provides substantive evidence that the reputation-policy paradigm is a welcome addition to the Grid computing community.