Using Step-Wise Refinement to Build a Flexible Lightweight Storage Manager

In recent years the deployment of embedded systems has increased dramatically, e.g. in the domains of sensor networks or ubiquitous computing. At the same time the amount of data that have to be managed by embedded systems is growing rapidly. For this reason an adequate data management support is urgently needed. Current database technologies are not able to cope with the requirements specific to embedded environments. Especially the extreme resource constraints and the diversity of hardware plattforms and operating systems are challenging. To overcome this tension we argue that embedded database functionality has to be tailored to the application scenario as well as to the target platform. This reduces the resource consumption and customizes the data management to the characteristices of the plattform and the application scenarion. We show that component techniques and feature-oriented programming help to face the mentioned limitations without focusing on special-purpose software. We present the design and the implementation of a database storage manager family. We discuss how feature-oriented domain analysis and feature-oriented programming help to do this task. Our evaluation criteria are the number of features and the flexibility to combine these features in different valid variants.