Database Query Processing in Main Memory

本研究项目的目标是开发新的查询评估和优化技术，用于处理主存数据库中的关系查询。最近在主存大小和成本方面的改进表明，大型应用程序可以通过将数据存储在主存中获得性能优势。该项目侧重于计算机体系结构中影响主存性能的问题，包括缓存丢失延迟和分支错误预测惩罚。数据库操作的算法将被开发出来，这些算法对这些问题和其他问题很敏感，因此在现代体系结构上表现良好。更广泛的问题，如如何设计一个全面的体系结构敏感的查询处理框架将被研究。这个项目产生的算法和技术可以应用于商业和实验数据库系统，在那里它们可以提高查询处理的速度。该项目的成果将作为研究论文和免费提供的原型软件分发。

The goal of this research project is to develop new query evaluation and optimization techniques for processing relational queries in main-memory databases. Recent improvements in main memory size and cost suggest that large classes of applications may obtain the performance benefits of having the data in main memory. The project focuses on issues in computer architecture that influence main memory performance, including cache miss latency and branch misprediction penalties. Algorithms for database operations will be developed that are sensitive to these and other issues, and thus perform well on modern architectures. Broader questions such as how to design a comprehensive architecture-sensitive query processing framework will be studied. The algorithms and techniques resulting from this project could have application in commercial and experimental database systems, where they could improve the speed of query processing. Results from the project will be disseminated as research papers and as freely available prototype software.