High-performance XML storage/retrieval system

This paper describes a system that integrates full-text searching and datab ase technologies for storing XML (eXtensible Markup Language) documents and retrieving information from them while providing a uniform interface. Our main goal with this system is to achieve high-performance, because there wi ll be a large amount of XML documents in the near future if XML becomes a s tandard for structured documents and data exchange. We have therefore devel oped techniques for achieving highperformance storage and retrieval of XML documents. For full-text searches, we improved the Structure Index + Text I ndex model, which references both indexes alternately at retrieval. In our improved method, a hieriarchical structure query is converted into a flat s tructure query by referencing just the structure index, then the optimized query can be quickly processed using only the text index. For storage, we d eveloped an offset space, which is an address space in secondary memory tha t can compactly store any structure, for example, a tree. We use the offset space to solve the problem that occurs in other methods which store the an alyzed result of XML documents as multiple relations in an nos. In our meth od, the analyzed result can be stored in a single page in the best case. Th is makes it superior to other methods which store the analysis results in m ultiple relations so that storage of N relations needs at least N pages. As a result, generally, our method greatly reduces I/O costs.