ORM Optimization through Automatic Prefetching in WebDSL

Abstract

Object-Relational Mapping (ORM) frameworks can be used to fetch entities from a relational database. The entities that are referenced through properties are normally not fetched initially, instead they are fetched automatically by the ORM framework, when they are used by the application. This is called lazy-fetching and can result in many queries, causing overhead. The number of queries can be reduced by prefetching multiple entities at once. There are two types of prefetching techniques, static and dynamic. Static techniques perform optimization during compilation and dynamic techniques collect information during runtime in order to perform prefetching. Multiple static prefetching techniques are implemented into WebDSL that all use the same static code analysis, however, they generate different queries. The static analysis determines the entities that are going to be used and should be prefetched. These static techniques are compared to the dynamic techniques already present inside the Hibernate ORM framework. The evaluation is performed using the OO7 benchmark and complete WebDSL applications. The results of the OO7 benchmark show a response time improvement of up to 69% over lazy-fetching. On complete web applications some of the static techniques implemented in WebDSL improve the performance on average, however, the performance may be improved further, using a more fine-grained method of choosing an optimization technique.