Research On Some Key Technologies Of Service Oriented Data Integration

With the rapid development of service computing in Internet and enterprise information network, as well as B2B,O2O, LBSNS, B2A and other new service models are emerging,the cooperation and interaction between the services is explosive growth in the service network. At the same time, the characteristics of the service network such as autonomy, loose coupling, and weak supervision lead to a sharp rise in the complexity of data services, Different format and content of the data service make the problem such as high data integration cost and low resource utilization more prominent.Data service uses the idea of service oriented and the data is encapsulated as a service. In composite services, the providing process of data service capability is also the service process of traditional data integration mode. Compared with the traditional method of federated database and data warehouse, the service oriented data integration has the advantages of flexible expansion, dynamic resource scheduling and integrated real-time,which is suitable for the network environment of complex data services.Based on service computing, the paper studies the integration of data services in the process of service composition,Research on four stages of data service processing, that is, service discovery, service selection, service scheduling and service composition,establish a service composition framework based on anomaly perception, design the service discovery method based on the comprehensive similarity, and establish the data service selection strategy based on trust degree, so as to realize the effective processing of the data service in the composite service, and guarantee the overall quality of the service system.The main innovation of this paper is:1) Aiming at the problem of low matching accuracy of large amounts of homogeneous and heterogeneous data sources,a new matching algorithm based on comprehensive similarity named ComSM (Comprehensive Similarity Matching) is proposed, match the demand for services and publishing service from three aspects of data, functional and non functional properties, in the calculation of similarity data, escape data service through semi-structured document reconstruction, analyse data service similarity from two aspects of structure and content, at the same time, two methods of describing attributes and semantics are used to compute the similarity of service function.In addition, taking into account the service quality of service users, the selection strategy based on K means algorithm is designed, and the matching of the three dimensions is deduced from the service candidate set.The experimental comparison shows that the accuracy of the comprehensive similarity matching algorithm is improved by 18% compared with the property of the hybrid algorithm.2) In order to evaluate the credibility of service under the integrated environment of large scale data services, a trusted service selection algorithm TLVCSS(Three Layer Validation Credible Service Selection) based on three layers is proposed, the algorithm considers the relationship between the trust degree of the service and the service dependency. The algorithm satisfies the requirement of the data service, and promotes the accuracy of the service selection through the mutual restriction between the data service.Firstly, the data characteristics are integrated into the degree of trust, and a data service quality trust model is established,use a time division evaluate method to compare the service demand QoS and service QoS continuously, to improve the accuracy of the candidate service set.Then, the service context is used to select the trusted service. Finally, the quality of data service is considered as the factor to optimize the service data. Experiments show that in the process of data service selection, compared with the traditional trusted service selection algorithm, this algorithm saves about 12% of the published service QoS calculation, and improves the accuracy of the service selection of about 23%.3) In view of the problem of service replacement for G3C Cost Constraints (Compensation), a service scheduling algorithm based on global compensation cost constraint is proposed.The algorithm first analyzes the QoS changes in the long transaction model and the effect of the data quality on the compensation operation.and realize the balance of the local compensation and the global compensation;Then, the dynamic mechanism is proposed to improve the execution efficiency of the compensation operation. Finally, the content of the global compensation price is improved by combining the loss of the replacement cost. In the experiment, the success rate of the compensation mechanism of the dynamic mechanism is improved by about 32% compared with the traditional compensation mechanism.4) Aiming at the problem that the accuracy of service composition update is reduced when the data integration services middleware is abnormal, a data service composition algorithm EPDSP(Environmental Perception Data Service Portfolio) based on environment aware is proposed,and the framework of service composition LPSC(Lord; Passive Abnormal Perception of Service Composition) is proposed, which is based on active and passive.abnormality perception.Firstly, an abnormal behavior model of active and passive composite services is proposed,and use the model,monitoring the main and passive service abnormalities. Subsequently, by the EPDSP algorithm, use heuristic algorithm to divide the self-healing candidate set, then self healing and update existing services through computing resources, data resources and network resources,Based on this, proposes a perception based composite service framework LPSC, which integrates the anomaly detection module, service discovery module, service training and service resource allocation module.and to improve the self-healing ability of the composite service by means of combining the combination model.