Research-oriented Data Manage System In Numerical Prediction

The numerical prediction of the weather is an important research area in the present world. With the progress of the climate science, the numerical prediction theory and the computer technique, especially the evolution of SOA, the numerical prediction turns its way from the conventional script style to workflow based on SOA.The way of workflow based SOA needs the support of the data integration and access urgently.Data integration technique is a way to dealing with data island, and an effective approach to solve the problem of sharing and access data in a distribute environment. Traditional data integration focuses on sharing and accessing disparate data resource with a seamless and transparent way.Due to its limits of processing automation, traditional data integration cannot satisfy the need of numerical prediction based SOA.Considering the requirement of data integration and acess in the numerical prediction based SOA, based on the research of the integration concept and related techniques, this thesis explores the difference of data integration in traditional and under SOA.In this process,this thesis presents a approach to store XML metadata in relational database that exploits the specific characteristics of a metadata catalog,and later establishes a plug-in based framework named spider that crawls the climate data and crosswalks metadata,and then derives a data service integrated by the relative data services in Globus Toolkit.In the end,based on spider framework and the integrated data service,this thesis applies the the XML-Relation storing and accessing technique mentioned above to realizing the climate data integration and access system in the numerical prediction environment on SOA.