2009.02.24 | Dorthe Haagen Nielsen

Infrastructure Support for Collaborative Pervasive Computing Systems

Abstract

Collaborative Pervasive Computing Systems (CPCS) are currently being deployed to support areas such as clinical work, emergency situations, education, ad-hoc meetings, and other areas involving information sharing and collaboration.These systems allow the users to work together synchronously, but from differentplaces, by sharing information and coordinating activities. Several researchers have shown the value of such distributed collaborative systems.

However, building these systems is by no means a trivial task and introduces a lot of yet unanswered questions. The aforementioned areas, are all characterized by unstable, volatile environments, either due to the underlying components changing or the nomadic work habits of users. A major challenge, for the creators of collaborative pervasive computing systems, is the construction of infrastructures supporting the system.

The complexity of building a distributed system, even in a controlled environment, is challenging. Introducing mobile and pervasive environments, with unstable hosts and links or freely moving users, multiply the complexity by factors. The creator has to implement complex software architectures and distribution mechanisms in order to build a supporting infrastructure for the envisioned collaborative pervasive computing system.

Following an experimental computer science approach, where experiments are used both for theory testing and for exploration, we research the area of infrastructure support for collaborative pervasive computing systems. After identifying ?ve main challenges for infrastructures for CPCS, we address these challenges from three different angles, namely, software architectures, programming paradigms, and distribution methods. This research leads us to present four major contributions to the area.

Firstly, we contribute by building real world Collaborative Pervasive Computing Systems, including the Activity-Based Collaboration system and the iHospital system, which has beendeployed and evaluated.
Secondly, we contribute with novel hybrid and fusion Software Architectures. Moreover, we propose separating the event ?ow in a distributed collaborative system, based on the nature of the single event. Thirdly, we contribute with peer-to-peer distributed shared objects as a Programming Paradigm and basic distribution mechanism. We have created and evaluated a framework called DOLCLAN as a full-blown implementation of the proposed concepts. Finally, we contribute to the Distribution Mechanismsresearching epidemic protocols or gossiping as distribution protocols for pervasive environments.

Presenting thesefour contributions to the area of infrastructure support for CPCS, this research contributes to an area, that holdsgreat possibilities for changing the way we build collaborative pervasive computing systems.