Ontology-based Mpi + Openmp Parallel Retrieval Applications

With the requirement for computing performance is more and more high in the fields of science and engineering, high-performance computer and cluster system are used more and more widely, especially the development of multi-core processors greatly improves the performance of parallel computing. This paper studies the hybrid programming model according to the structural characteristics of multi-core cluster, and applied it to the Chinese medicine case analysis system.Firstly, this paper researches and analyses the parallel computing theory and architecture, especially the multi-core architecture. Multi-core processor is a processor integrated two or more complete computing cores, compared with the single-core processor it has higher computing power. How to take full advantage of the characteristics of multi-core cluster systems to improve the speedup and the resource utilization is the focus and the hot spots of the study today. This paper studies the MPI+OpenMP hybrid programming model suitable for the multi-core cluster system combine with the hierarchical structure of multi-core system, and achieves the process and thread two-level parallel base this. MPI+OpenMP hybrid programming model is scalable, portable, easy to realize, etc, takes full play to the advantages of cluster system.Then, this paper research related technology of semantic retrieval, analyses the ontology and the semantic extension. The semantic retrieval assigns semantic to key words, expands the upper and lower words, synonyms and synonyms, increases the accuracy and recall. This paper describes the semantics by ontology, uses the OWL language as the ontology description language. According to establish the ontology database between users and data resource, applies semantic inference, designs ontology-based semantic extension retrieval framework.Finally, this paper analyses the ontology-based semantic retrieval technology, studies ontology-based parallel retrieval technology, and realizes the parallel of this model in 3 steps:(1) The parallel of ontology database structure and the parallel between nodes; (2) The parallel of reasoning process and thread-level parallel on the master node;(3) The parallel of retrieval, the retrieval conditions that are semantic extended are distributed to each node to realize the parallel between nodes.In this paper, ontology-based parallel retrieval model is applied to the retrieval model of Chinese case analysis system, and is run on multi-core blade cluster. According to the performance analysis of the results, we can see:ontology-based search model that studied in this paper can take advantage of parallel cluster structure and improve the retrieval speed.