Geographical distribution schemes to support Publish/Subscribe in Mobile Ad Hoc Networks

Mobile Ad Hoc Networks (MANETs) allow communication between moving nodes which does not depend on infrastructure like access points, stationary routers or GPS satellites. This offers new communication opportunities, especially in challenging environments. To communicate in MANETs we often need routing functionality, which usually provides unicastbased functionality. Publish/Subscribe (Pub/Sub) is a well known and powerful paradigm that provides higher expressiveness than unicast routing. It decouples senders from receivers and allows information exchange between network nodes that offer certain data (called publishers) and nodes that declare their interest in data of some pattern (called subscribers). Especially in MANET applications, Pub/Sub provides functionality useful to support realistic scenarios and novel applications. In most cases Pub/Sub applications are used in wired networks. The system models of wired networks are different to MANETs, hence solutions for wired networks are not feasible in mobile context. This diploma thesis investigates which concepts and distribution schemes are feasible to support Pub/Sub in MANETs in an efficient way. A new algorithm (TPSR) is proposed, tailored to efficiently manage data distribution to support Pub/Sub in MANETs. TPSR contains two innovative principles. First it uses the dissemination of subscriptions to create source routes. Second it makes use of the signal strength messages are received with, to optimize routes in terms of striking a good balance between long routes and fragile routes. The simulator Ns-2 is used to evaluate the different solutions and to discuss problems and solutions, which heavily depend on the scenarios used. Acknowledgements I thank the DAAD (Deutscher Akademischer Austausch Dienst) for funding two research journeys to Lisbon, Portugal (Project 0813002). I also thank my supervisor Prof. Kurt Rothermel for his support and help on choosing the topic of this thesis. Special thanks goes to my two advisors, Dr. Boris Koldehofe and Dr. Hugo Miranda for teaching me how research works. They guided me during the whole 6 months. Especially during the last week they supported me a lot to get my thesis finished in time, providing valuable comments and suggestions. For providing infrastructure, computing power and related work I also thank the Distributed Systems Department of the University of Stuttgart. Beside human help, I thank God my father for his endless love and support during the past six months.