Database design and tuning for XML workloads

XML已经成为表示半结构化数据和在应用程序之间交换消息的事实标准。很长一段时间以来，使用XML数据的应用程序可以从将数据存储在数据库系统中并以与处理关系数据类似的方式处理数据的能力中受益匪浅。然而，XML数据具有许多与关系数据截然不同的特征，这意味着关系数据库系统不能为XML数据提供足够的支持。这导致了大量旨在为XML数据库提供支持的学术和商业研究项目。在XML数据库中，就像在关系数据库中一样，数据库的物理设计在决定数据库在给定工作负载下的性能方面起着关键作用。数据库的物理设计指定构造哪些辅助结构(索引和物化视图)、如何对数据进行分区以及如何在不同的节点和存储设备上布局数据。在关系数据库中，已经对开发工具和技术进行了大量研究，以帮助数据库管理员完成困难但重要的物理数据库设计任务。对于XML数据的物理数据库设计问题，人们知之甚少，在数据库研究界也没有得到充分的解决。该项目将解决这一问题，目标是提供工具和技术，用于自动化针对XML数据和工作负载的物理数据库设计。

XML has become the de facto standard for representing semi-structured data and for exchanging messages between applications. It has been clear for a long time that applications that use XML data could greatly benefit from the ability to store this data in database systems and manipulate it in ways similar to the ways that they manipulate relational data. However, XML data has many characteristics that are quite different from relational data, which means that relational database systems are not capable of providing adequate support for XML data. This has led to a large number of academic and commercial research projects that aim at providing support for XML databases. In XML databases, just as in relational databases, the physical design of the database plays a key role in determining the performance of the database for a given workload. The physical design of a database specifies which auxiliary structures (indexes and materialized views) are constructed, how the data is partitioned, and how it is laid out on the different nodes and storage devices. In relational databases, there has been a substantial amount of research into developing tools and techniques for helping database administrators with the difficult but important task of physical database design. The problem of physical database design for XML data is much less understood, and has not been adequately addressed in the database research community. This project will address this problem, with the goal of providing tools and techniques for automating physical database design for XML data and workloads.