Research On Component Retrieval Of Reconfigurable Routing Switching Platform

Three networks convergence is regarded as the future trend of network development. The key equipment of three networks convergence is open architectonic reconfigurable routing switching platform, which is supported by the platform with component-based processing technologies. Component management is implemented by component library. A perfect and efficient library is the crucial infrastructure of a systematized and practical platform. The features of a component library in a reconfigurable routing switching platform are as follows: a huge number of changing components, and a high efficiency required by service. However, the current component retrieval algorithms can't satisfy the managing of dynamic changes of component library, and also are insufficient in processing massive components.Combined with the fundamental technique research task of the"Research of reconfigurable router and its components"project of the National High-Tech Research and Development Program of China (863 Program), this dissertation mainly studies the component retrieval algorithm as well as engineering implementation of retrieval the massive and dynamically changing components of a library in order to improve the low efficiency of the current retrieval algorithm. The major works of this dissertation are as follows:(1) In order to tackle the problem that the general classification cannot present the features of routing, a routing component classification method based on facet is presented, constructs a model and tree scheme of facet classification. The classification method lays a solid foundation for the following component retrieval algorithm.(2) As it is insufficient for the traditional component retrieval mechanism in managing the dynamic changes of component library, a component retrieval algorithm based on facet tree code matching (FTCM) is proposed. In reference of the widely used retrieval algorithm based on tree matching, the new algorithm transfers the problem of tree matching to string matching, lower the time complexity, improving the ability of managing the dynamic changes of the component library.(3) In order to present the matching relationship between changing components in library and the user requirements, this dissertation add the result of interactive calculation of genetic algorithm to the facet classification tree, propose a component retrieval algorithm based on facet tree code matching of dynamically reconfigurable weight (FTCMDRW). The value of the algorithm can change dynamically according to the component library and user requirements, further improving the ability of processing the fluctuate library; according to the matching degree, this algorithm can put the retrieved components in order, insure the recall and increase the precision at the same time. Results of emulation experiments reveal that after the dynamically reconfigured service requirement accepted by the platform, the new algorithm has a 12% better precision than the conventional ones.(4) In order to improve the retrieval mechanisms'ability of processing massive components, a component retrieval algorithm based on fuzzy inference-back propagation neutral networks (FI-BPNN) is proposed, by using the parallel processing ability of the neural network and the easy implementation of fuzzy inference. Stimulation tests reveal that this algorithm is especially efficient with complex retrieval problem in component library. When the number of components reaches 1000 or more, the average precision is 14%better.(5) To meet the requirements of the subject, this dissertation designs and implements a reconfigurable routing switching platform component library prototype system, brings the retrieving module into function and puts this algorithm into practice. With the data from experiments on this prototype system, further demonstration on the feasibility and efficiency of this retrieval algorithm are also referenced.