User-Defined Aggregates for Advanced Database Applications

信息服务的规模和复杂性的爆炸性增长正在使数据库技术超出其极限。特别是，虽然最先进的数据库管理系统(DBMS)通过用户定义函数(UDF)和数据类型提供了可扩展性，但它们在许多关键应用领域(例如数据挖掘)是无效的。因此，该项目的目标是开发强大的可扩展性机制，使DBMS能够有效地支持新的应用领域和高级信息系统。为了实现这一目标，新方法基于以一种新的基于SQL的语言表示的用户定义聚合(UDA)。因为SQL是DBMS的标准语言，所以这种方法确保了易用性和兼容性--克服了困扰基于UDF的方法的问题。该项目的预期成果是：(1)为UDA设计和实施一种基于SQL的语言，以及支持其有效执行的系统；开发编译和优化技术，以确保UDA在大型数据库管理系统驻留数据库以及大型文件驻留数据上的性能和可扩展性；(2)开发高级应用试验台，以展示UDA系统的能力，并评估和调整其性能；在大多数应用中，如数据挖掘功能、时空查询和时间序列数据刀片，该系统将用作数据库管理系统的扩展器；但在应用程序中，如分析计算数据或网络数据，系统将被用作独立的查询处理器；以及(3)向公众交付软件，并有一个专门的网站提供文档、教程、应用程序试验台和在线demos.http://www.cs.ucla.edu/~edu/udahome.html

An explosive growth in scale and complexity of information services is stretching data base technology beyond its limits. In particular, while the state-of-the-art data base management systems (DBMSs) provide extensibility through user-defined functions (UDFs) and data types, they are ineffective in many critical applications areas, e.g., data mining. Therefore, the project's goal is developing powerful extensibility mechanisms to enable DBMSs to support effectively new application domains and advanced information systems. To realize this goal, the new approach is based on User-Defined Aggregates (UDAs) expressed in a new SQL-based language. Since SQL is the standard language for DBMSs, this approach ensures ease of use and compatibility -- overcoming a problem besetting UDF-based approaches. The project's expected accomplishments are: (1) Design and implementation of an SQL-based language for UDAs, and of a system for supporting their efficient execution; development of compilation and optimization techniques that ensure UDAs' performance and scalability on large DBMS-resident databases, as well as on large file-resident data; (2) Development of advanced application testbeds to demonstrate the power of the UDA system, and to evaluate and tune its performance; in most applications, such as datamining functions, spatio-temporal queries, and time series data blades, the system will be used as a DBMS extender; but in applications, such as analysis of computational data or WEB data, the system will be used as a stand-alone query processor; and (3) Delivery of the software to the public, with a dedicated website providing documentation, tutorials, application testbeds, and online demos.http://www.cs.ucla.edu/~edu/udahome.html