Since XML emerged as a standard of data exchange and representation on the web, there has been extensive research conducted on XML data processing and management. In the near future, we expect the era of ubiquitous computing where the mobile devices s...
Since XML emerged as a standard of data exchange and representation on the web, there has been extensive research conducted on XML data processing and management. In the near future, we expect the era of ubiquitous computing where the mobile devices such as smart phones play a key role in full blossom, and we still need to investigate into more effective XML query processing for such an environment because the mobile devices are limited in resources.
In this thesis, we describe research on XML query processing when XML document is stored in the database server, and user queries in XPath need to be processed in a more efficient way. First of all, we examine background knowledge and previous work. Among others we focus on tag name-based query processing which is now regarded as the most representative approach to XML query processing, and analyze the problems inherent in that approach. It has the following two major limitations: First, its query tree consists of tag name-labeled nodes. As such, when the XML node list for each query tree node is retrieved from the XML database, the retrieval is through tag name match. Thus, for a query tree node, all the irrelevant XML nodes with the same tag name but not on the target path are retrieved as well. Secondly, though only a subset of the query tree nodes needs to be dealt with in producing a correct query result, every query tree node is accessed for structural joins that check the parent-child or ancestor-descendant relationship. As such, query performance is degraded.
To overcome these problems, we investigate into path-based query processing using the reduced query tree. First, instead of the conventional tag name-based query tree, we employ its path-based reduced query tree to eliminate unnecessary structural joins. Secondly, when the XML node list is retrieved, we conduct path identifier match instead of tag name match, avoiding the retrieval of the XML nodes irrelevant to the query. This approach results in a significant improvement in query performance in most of the queries. Its overhead is the transformation of the tag name-based query tree into its reduced counterpart, and the storage space for the tag structure of the XML document. Through implementation and extensive set of experiments, we show that the path-based approach outperforms the tag name-based one and that the aforementioned overhead is negligible.