Extensible Markup Language (XML) is a de facto standard for data exchange in the World Wide Web. Indexing plays a key role in improving the execution of XML queries over that data. In this thesis we discuss the three main categories of indexes proposed in the literature to handle the XML semistructured data model, and identify limitations and open problems related to these indexing schemes. Based on our findings, we propose two novel XML index structures to overcome most of these limitations: a native index structure called Level-based Tree Index for XML databases (LTIX) and a relational index structure called Universal Index Structure for XML data (UISX). A proper labeling scheme is an essential part of a well-built XML index structure. We found that existing labeling schemes are not suitable for our index structures and therefore propose a novel labeling scheme, Level-based Labeling Scheme (LLS), which has the advantages of most popular types of labeling schemes while eliminating the main disadvantages. We then combine our LLS labeling scheme with our index structures. An evaluation shows that LLS performs well in comparison to existing labeling schemes using different mappings to relational tables. We propose the LTIX to minimize the number of joins and matches required to evaluate twig queries, and also to facilitate effective query optimization through early pruning of the space search. Our experimental results show that this approach performs well in comparison to existing state-of-the-art approaches. We propose the UISX to overcome the key problem with the state-of-the-art approaches, namely that they cannot support efficient processing of twig queries without requiring significant storage. We use a light-weight native XML engine on top of an SQL engine to perform the optimization related to the structure of the XML data prior to shredding. Experimental results show that our approach achieves lower response times than other similar approaches while using less space to store XML data. / Thesis (Ph.D, Computing) -- Queen's University, 2011-03-15 23:03:50.15

Database systems employ physical structures such as indexes and materialized views to improve query performance, potentially by orders of magnitude. It is therefore important for a database administrator to choose the appropriate configuration of these physical structures (i.e., the appropriate physical design) for a given database. Deciding on the physical design of a database is not an easy task, and a considerable amount of research exists on automatic physical design tools for relational databases. Recently, XML database systems are increasingly being used for managing highly structured XML data, and support for XML data is being added to commercial relational database systems. This raises the important question of how to choose the appropriate physical design (i.e., the appropriate set of physical structures) for an XML database. Relational automatic physical design tools are not adequate, so new research is needed in this area. In this thesis, we address the problem of automatic physical design for XML databases, which is the process of automatically selecting the best set of physical structures for a given database and a given query workload representing the client application's usage patterns of this data. We focus on recommending two types of physical structures: XML indexes and relational materialized views of XML data. For each of these structures, we study the recommendation process and present a design advisor that automatically recommends a configuration of physical structures given an XML database and a workload of XML queries. The recommendation process is divided into four main phases: (1) enumerating candidate physical structures, (2) generalizing candidate structures in order to generate more candidates that are useful to queries that are not seen in the given workload but similar to the workload queries, (3) estimating the benefit of various candidate structures, and (4) selecting the best set of candidate structures for the given database and workload. We present a design advisor for recommending XML indexes, one for recommending materialized views, and an integrated design advisor that recommends both indexes and materialized views. A key characteristic of our advisors is that they are tightly coupled with the query optimizer of the database system, and rely on the optimizer for enumerating and evaluating physical designs whenever possible. This characteristic makes our techniques suitable for any database system that complies with a set of minimum requirements listed within the thesis. We have implemented the index, materialized view, and integrated advisors in a prototype version of IBM DB2 V9, which supports both relational and XML data, and we experimentally demonstrate the effectiveness of their recommendations using this implementation.

Database systems employ physical structures such as indexes and materialized views to improve query performance, potentially by orders of magnitude. It is therefore important for a database administrator to choose the appropriate configuration of these physical structures (i.e., the appropriate physical design) for a given database. Deciding on the physical design of a database is not an easy task, and a considerable amount of research exists on automatic physical design tools for relational databases. Recently, XML database systems are increasingly being used for managing highly structured XML data, and support for XML data is being added to commercial relational database systems. This raises the important question of how to choose the appropriate physical design (i.e., the appropriate set of physical structures) for an XML database. Relational automatic physical design tools are not adequate, so new research is needed in this area. In this thesis, we address the problem of automatic physical design for XML databases, which is the process of automatically selecting the best set of physical structures for a given database and a given query workload representing the client application's usage patterns of this data. We focus on recommending two types of physical structures: XML indexes and relational materialized views of XML data. For each of these structures, we study the recommendation process and present a design advisor that automatically recommends a configuration of physical structures given an XML database and a workload of XML queries. The recommendation process is divided into four main phases: (1) enumerating candidate physical structures, (2) generalizing candidate structures in order to generate more candidates that are useful to queries that are not seen in the given workload but similar to the workload queries, (3) estimating the benefit of various candidate structures, and (4) selecting the best set of candidate structures for the given database and workload. We present a design advisor for recommending XML indexes, one for recommending materialized views, and an integrated design advisor that recommends both indexes and materialized views. A key characteristic of our advisors is that they are tightly coupled with the query optimizer of the database system, and rely on the optimizer for enumerating and evaluating physical designs whenever possible. This characteristic makes our techniques suitable for any database system that complies with a set of minimum requirements listed within the thesis. We have implemented the index, materialized view, and integrated advisors in a prototype version of IBM DB2 V9, which supports both relational and XML data, and we experimentally demonstrate the effectiveness of their recommendations using this implementation.