Research On Key Techniques Of Trust Evaluation Oriented Resource Matching In VCE

On the basis of open network infrastructure, virtual computing environment (VCE) creates a ubiquitous and intelligent public computing platform by integrating the computing, storage and network resources in the internet. As a new form of realization, VCE aims at providing users with a convenient, fast and effective application environment, and supporting the application with integrated services that are harmonious,efficient and reliable. Unlike the traditional computing environment, the public environment faces the challenges of disorderly growth, high autonomy and heterogeneity of network resources, and the complexity,unpredictability and diversity of a massive amount of applications. These challenging factors directly impact the accurate matching between the virtual resources and the application tasks, and constrain the provision of trusted services in the public environment.Targeted at the resource matching issue in VCE and in view of the running of the integrated experiment platform, this paper studies the key technologies on trust evaluation for resource matching in public environment by constructing an open trust evaluation mechanism, and performing cluster analysis of virtual resources and tasks before the trust evaluation. A series of innovative research results with theoretical and practical value has been achieved. Specifically, major research and innovative results are demonstrated in the following four aspects:(1) Presenting a trust evaluation framework for resource matching in VCE. Firstly, considering the basic theories and structures of VCE, this paper digs deep into the demand for trust in resource matching; Secondly,based on the theory of scientific evaluation and trust management, this paper specifies that the integrated evaluation should consist of the steps of identifying evaluation targets, selecting evaluation objects, establishing evaluation indices, and determining evaluation weight, and sets up an integrated evaluation model based on the hierarchical structure formed by the AHP during weight setting of evaluation indices; Finally, in respect to resource matching objects, i.e. online registration and online operation of virtual resources and tasks, this paper establishes a trust evaluation mechanism from multiple levels, including the identity, competence,action of virtual resources and the identity, demand and consumption of the tasks, and puts forward the research method of putting cluster analysis before trust evaluation. Thus, this paper lays the theoretical foundation for the provision of efficient and trusted integrated services and provides solid theoretical basis for follow-up study of resource matching optimization.(2) Presenting a clustering model based on resource matching parameters. Designed to maintain the precision and efficiency of resource matching, the model utilizes machine learning theory to explore the effect of the multi-dimension parameters of virtual resources and tasks on resource matching. Extracting features of resource allocation, service quality and actual resource consumption, the service-driven algorithm acquires the feature attributes related to virtual resources or tasks, and analyzes the virtual resources and tasks in different clusters. In task clustering, the tasks under different demands are divided into different clusters based on parameters like task demand, consumption and other feature attributes, pre-treated by category, and subject to cluster analysis based on the improved DBSCAN algorithm; In consideration of the constraints of resource load on resource matching, the clustering model of virtual resources realizes the clustering analysis of virtual resources based on the improved bisecting K-means algorithm. The clustering model contributes to the improvement of resource matching efficiency and precision, and boosts the applicability of virtual computing experiment platform(3) Presenting a trust evaluation model for virtual resources and tasks. Starting from the trust evaluation framework for resource matching in VCE, this paper presents such a model which integrates various layers of trust evaluation in the whole cycle of virtual resources and tasks. This model carries out evaluation from multiple angles, ranging from identity trust, competence trust, action trust, demand trust to consumption trust,and calculates the credibility with functions like time decay function,reward function, etc. Plus, this model works effectively under inveracious service, external inveracious service or malicious service, thereby ensuring a rate of successful transactions and further promoting the practicality of resource matching.(4) Presenting a resource matching optimization strategy based on clustering and trust evaluation in VCE. Resource matching is not only the basis for realizing on-demand resource aggregation and autonomous resource coordination mechanism, but also the key to realize a string of goals from improving the service quality and resource utilization rate to ensuring the service credibility. According to the resource matching optimization strategy, the virtual resources and tasks are pre-classified and subject to cluster analysis,forming a categorical search range for resource matching. In other words, the trust evaluation results of virtual resources and tasks are acquired by category to achieve swift and accurate resource matching. Besides, the strategy also optimizes the success rate and completion time of the tasks after resource matching.Suffice it to say that the strategy boasts significant practical value and broad application prospects.