Distributed Persistence for Limited Devices

The problem of storing data both locally and remotely in a synchronized fashion is common in ubiquitous computing, where a mobile device (e.g., a tiny sensor node or a mobile phone) might produce a significant amount of data while having only limited storage capacity. While data management frameworks for small and mobile devices exist, they either have high system requirements (e.g., JavaSpaces) or do not focus on distributed storage issues (e.g., TinyDB). We have designed and implemented a lightweight storage system that allows the transparent use of both local and remote storage space, using identical semantics. It is based on a serialization framework that allows CLDCenabled J2ME devices to transparently store Java objects in the (local) Record Management Store, or in a remote database via a wireless byte stream transmission. We have built a prototypical mobile application ontop of this framework and report on our experiences.