CAREER: Generating Application-Specific Database Management Systems

数据库管理系统(DBMS)被设计为支持从银行到社交网络和进行科学发现的各种应用程序的通用工具。为了提高这类应用程序的性能，研究人员利用应用程序领域的独特特征来构建性能优于传统实现的特定于域的DBMS。执行这种专业化需要劳动密集、复杂且容易出错的工作。该项目的智能优点是通过研究自动执行这种领域专门化的技术来提高特定于应用的DBMS设计的最新水平。作为该项目更广泛影响的一部分，所学到的经验和技术将被整合到PI常规教授的编程语言和课程中。具体地说，这项提议旨在利用编程系统和数据管理研究的最新进展来建立能够自动理解数据库应用程序语义的工具。有了这样的知识，这个项目的目标是1)创建能够自动优化应用程序可能发出的特定查询集的工具，并证明优化后的查询在语义上与输入等价；2)研究自动选择最佳框架(在执行时间、所需资源等方面)以执行应用程序发出的查询的技术；3)为程序员设计新的语言，以在跨分布式节点集执行查询时表达他们的数据一致性需求，并构建这些语言的实现。在这个项目中开发的所有软件构件都向公众发布，并计划将它们的使用纳入本科生和研究生的课程。此外，作为收集和研究现实世界数据库应用程序缺点的项目的一部分，收集的应用程序被收集到一个储存库中，该储存库可供外地的研究人员和从业人员进行实验和复制结果。

Database management systems (DBMSs) are designed to be general-purpose tools that support a wide variety of applications, from banking to social networking and making scientific discoveries. To improve the performance of such applications, researchers have leveraged the unique characteristics of application areas to build domain-specific DBMSs that outperform traditional implementations. Performing such specialization requires labor intensive, complex, and error prone efforts. The intellectual merits of this project are to advance the state of the art in application-specific DBMS design by investigating techniques to perform such domain specialization automatically. As part of this project's broader impacts, the lessons and techniques learned will be integrated into programming languages and classes that the PI routinely teaches.Specifically, this proposal aims to leverage recent advances in programming systems and data management research to build tools that can automatically understand database application semantics. Given such knowledge, the goals of this project are to 1) create tools that can automatically optimize the specific set of queries that can potentially be issued by the application, and prove that the optimized queries are semantically equivalent to the inputs; 2) investigate techniques to automatically select the optimal framework (in terms of execution time, resources required, etc) to execute the queries issued by the application, and 3) devise new languages for programmers to express their data consistency needs when queries are to be executed across a distributed set of nodes, and build an implementation of such languages. All software artifacts developed in this project are released to the public, with plans to incorporate their usage in both the undergraduate and graduate curricula. In addition, as part of the project is to collect and study the shortcomings of real-world database applications, the collected applications are collected into a repository that is publicly accessible repository for researchers and practitioners in the field to experiment and reproduce the results.