Storage and Query Processing Optimizations for Hierarchically - Organized Data

Hierarchical data can be found anywhere multiple pieces of information are connected by a relationship. The first chapter of my thesis deals with processing relational queries in the context of a native XML storage system. We take advantage of hierarchical XML to equate the nested structure of the XML documents with the key relationship between two data items. A join query can then be rewritten as a storage scan, allowing acceleration of the query. The second chapter provides a comparison of row and column oriented storage optimizations given an assumption of read-mostly data access. We focus on two, “super tuples” and “column abstraction”, to elucidate the difference between row and column storage layouts. Column abstraction allows optimization of hierarchically organized data by storing repeating values only once. We extend the read-optimized relational store in the third chapter to evaluate two predicate evaluation strategies—a scan-based solution over both standard super tuples and the PAX layout, and an index-based strategy with and without a slot array in super tuples.