Deriving Efficient SQL Sequences via Read-Aheads

摘要

Modern information system architectures place applications in an application server and persistent objects in a relational database. In this setting, we consider the problem of improving application throughput; our proposed solution uses data prefetching (read-aheads) to minimize the total data-access time of an application, in a manner that affects neither the application code nor the backend DBMS. Our methodology is based on analyzing and automatically merging SQL queries to produce query sequences with low total response time, in ways that exploit the application's data-access patterns. The proposed approach is independent of the application domain and can be viewed as a component of container managed persistence that can be implemented in middleware. This paper describes our proposed framework for using generic data-access patterns to improve application throughput and reports preliminary experimental results on discovering key parameters that influence the trade-offs in producing efficient merged SQL queries. The approach is evaluated in the context of a financial domain, which yields the kinds of natural conceptual relationships where our approach is valuable.