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.

Experience with relational systems has shown that distribution is an effective way of improving the scalability of query evaluation. In this paper, we show how distributed query evaluation can be performed in a vertically partitioned XML database system. We propose a novel technique for constructing distributed execution plans that is independent of local query evaluation strategies. We then present a number of optimizations that allow us to further improve the performance of distributed query execution. Finally, we present a response time-based cost model that allows us to pick the best execution plan for a given query and database instance. Based on an implementation of our techniques within a native XML database system, we verify that our execution plans take advantage of the parallelism in a distributed system and that our cost model is effective at identifying the most advantageous plans.

While centralized query processing over collections of XML data stored at a single site is a well understood problem, centralized query evaluation techniques are inherently limited in their scalability when presented with large collections (or a single, large document) and heavy query workloads. In the context of relational query processing, similar scalability challenges have been overcome by partitioning data collections, distributing them across the sites of a distributed system, and then evaluating queries in a distributed fashion, usually in a way that ensures locality between (sub-)queries and their relevant data. This thesis presents a suite of query evaluation techniques for XML data that follow a similar approach to address the scalability problems encountered by XML query evaluation.Due to the significant differences in data and query models between relational and XML query processing, it is not possible to directly apply distributed query evaluation techniques designed for relational data to the XML scenario. Instead, new distributed query evaluation techniques need to be developed. Thus, in this thesis, an end-to-end solution to the scalability problems encountered by XML query processing is proposed.Based on a data partitioning model that supports both horizontal and vertical fragmentation steps (or any combination of the two), XML collections are fragmented and distributed across the sites of a distributed system. Then, a suite of distributed query evaluation strategies is proposed. These query evaluation techniques ensure locality between each fragment of the collection and the parts of the query corresponding to the data in this fragment. Special attention is paid to scalability and query performance, which is achieved by ensuring a high degree of parallelism during distributed query evaluation and by avoiding access to irrelevant portions of the data.For maximum flexibility, the suite of distributed query evaluation techniques proposed in this thesis provides several alternative approaches for evaluating a given query over a given distributed collection. Thus, to achieve the best performance, it is necessary to predict and compare the expected performance of each of these alternatives. In this work, this is accomplished through a query optimization technique based on a distribution-aware cost model. The same cost model is also used to fine-tune the way a collection is fragmented to the demands of the query workload evaluated over this collection. v To evaluate the performance impact of the distributed query evaluation techniques proposed in this thesis, the techniques were implemented within a production-quality XML database system. Based on this implementation, a thorough experimental evaluation was performed. The results of this evaluation confirm that the distributed query evaluation techniques introduced here lead to significant improvements in query performance and scalability both when compared to centralized techniques and when compared to existing distributed query evaluation tec...