Scaling XML query processing: distribution, localization and pruning

Abstract

Distributing data collections by fragmenting them is an effective way of improving the scalability of a database system. While the distribution of relational data is well understood, the unique characteristics of the XML data and query model present challenges that require different distribution techniques. In this paper, we show how XML data can be fragmented horizontally and vertically. Based on this, we propose solutions to two of the problems encountered in distributed query processing and optimization on XML data, namely localization and pruning. Localization takes a fragmentation-unaware query plan and converts it to a distributed query plan that can be executed at the sites that hold XML data fragments in a distributed system. We then show how the resulting distributed query plan can be pruned so that only those sites are accessed that can contribute to the query result. We demonstrate that our techniques can be integrated into a real-life XML database system and that they significantly improve the performance of distributed query execution.

Generating efficient execution plans for vertically partitioned XML databases

“…The construction of such execution plans is the main contribution of this paper and will be described in Section 3. For completeness, we briefly summarize the other steps in Appendix B; they are discussed in more detail elsewhere [15].…”

“…Label paths are useful not only for optimizing distributed plans but also for certain optimizations during query localization [15].…”

“…Due to XML's tree structure, an XML collection can be partitioned in a largely unconstrained fashion by cutting individual document edges (e.g., [1,9,11]). However, the concise fragmentation schema that can be obtained by fragmenting a collection based on characteristics of data or schema is a significant asset during distributed query optimization [4,7,15,16,17]. For our work, we have, therefore, chosen a distribution model that is based on schema characteristics and query operators (similar to relational distribution models): horizontal fragmentation based on selection, vertical fragmentation based on projection, and hybrid fragmentation based on a concatenation of both operators.…”

“…In order to perform distributed query evaluation (rather than straightforward data shipping), the query needs to be localized, yielding local query plans that can be evaluated at the sites holding the relevant data fragments. This process involves pruning unnecessary fragments to eliminate unnecessary work [11,15]. Finally, a distributed execution plan needs to be generated and optimized that assembles the results of local query plans into the overall query result.…”

“…• Based on our previous work on localization and local plan generation [15], we propose a technique for generating distributed execution plans within the context of a vertically partitioned and distributed XML database system. • We analyze the benefits and drawbacks of different types of execution plans.…”