Workload-aware trie indices for XML

Well-designed indices can dramatically improve query performance. Including query workload information can produce indices that yield better overall throughput while balancing the space and performance trade-off at the core of index design. In the context of XML, structural indices have proven to be particularly effective in supporting XPath queries by capturing the structural correlation between data components in an XML document. In this paper, we propose a family of novel workload-aware indices by taking advantage of the disk-based P[k]-Trie index framework, which indexes node pairs of an XML document to facilitate indexonly evaluation plans. Our indices are designed to be optimal for answering frequent path queries in one index lookup and efficient for answering non-frequent path queries using an index-only plan. Experimental results prove that our indices outperform the APEX index in overall throughput and excel in answering non-frequent queries, queries with predicates, and qu...

• Type: Conference
• Year: 2009
• Where: CIKM
• Authors: Yuqing Wu, Sofia Brenes, Hyungdae Yi
• Keywords: CIKM 2009, Database, Overall Throughput, Path Queries, Well-designed Indices