In dieser Arbeit werden P2P-Algorithmen in ressourcen-limitierten und irregulären Wireless-ad-hoc-Netzwerken (WAHN) betrachtet, die effizient, skalierbar und fehlertolerant in Situationen arbeiten sollen, in denen eine räumlich benachbarte Gruppe von Netzwerkknoten simultan ausfällt. Es wird ein fehlertolerantes Replikationsschema zur datenzentrischen Speicherung betrachtet, und eine selbstorganisierende, skalierbare Berechnung von Datenaggregaten zur Lösung des Konsensproblems. Existierende P2P-Algorithmen die Skalierbarkeit, Fehlertoleranz und Selbstorganisation in drahtgebundenen Netzen betrachten sind für die Klasse des WAHNs nicht geeignet weil sie Engpässe in WAHNs verursachen können und in Katastrophenmanagement-szenarien die Zuverlässigkeit der Daten nicht sicherstellen können. Die Verwendung von Informationen der geographischen Position von Knoten ist ein möglicher Weg, um die Effizienz und Skalierbarkeit von P2P-Anwendungen in drahtlosen Netzwerken zu verbessern. In dieser Arbeit wird ein neuer Ansatz vorgestellt, wie auf effiziente Weise 1) Gebiet des Netzwerks, das die geographische Ausbreitung seiner Knoten umfasst, und 2) Gruppenzugehörigkeit, wobei jeder Knoten zu genau einer Gruppe innerhalb eines einstellbaren Gebietes gehört, erzeugt werden kann. Dadurch können: existierenden, skalierbare P2P Datenspeicheralgorithmen für WAHNs genutzt werden, effiziente, fehlertolerante Replikation erstellt werden, die Effizienz von geographischen Routing und der Suche nach Replikaten verbessert werden sowie, Anwendungen auf einen bestimmten geographischen Bereich innerhalb des WAHN beschränkt werden (z.B. im Aggregationsprotokoll). Die entwickelten Protokolle sind tolerant gegenüber Nachrichtenverlust und verwenden ausschließlich lokale Broadcast-Nachrichten. Das Protokoll wurde mit Simulationen untersucht, die auf realistischen Netzwerktopologien mit Anteilen an sehr spärlichen und sehr dichten Knotenansammlungen basieren. / This dissertation addresses the challenge of reaching efficiency, scalability and fault-tolerance by P2P algorithms for resource-limited and irregular wireless ad-hoc networks (WAHNs) in disaster management (DM) scenarios where a spatially correlated group of nodes may crash simultaneously. In particular, we consider a fault-tolerant replication scheme for data-centric storage and a self-organized, scalable calculation of localized data aggregates for solving the consensus problem. Existing Peer-to-Peer algorithms that address issues of scalability, fault tolerance and self-organization in wired networks are inadequate for the addressed systems, they may cause bottlenecks in WAHNs and use replication that abstracts from geographical location of replicas and cannot therefore supply data survivability in DM scenarios in WAHNs. Incorporating information on geographical location of nodes is a recognized way to increase the efficiency and scalability of P2P applications in wireless networks. This dissertation proposes to efficiently construct new position information in a location-aware WAHN, where each node knows its own location and location of its direct neighbors. The new information are: network area, which expresses the geographical area covered by the network, and group membership, where each node belongs to exactly one group that is placed over the area of a maximum defined size. Together, they enable the use of the existing, scalable P2P data store in WAHNs (Geographical Hash Table), allow design of efficient fault-tolerant replication for the assumed fault model, increase efficiency of geographic routing and replica search, and allow to limit the geographical extent of activity of any distributed application, as we show using an example of data aggregation protocol. Proposed protocols tolerate message loss and use local broadcast only. They are evaluated by simulation over irregular topologies following the node placement of the existing, large WAHNs.

Mobility-Based Routing Overhead Management in Reconfigurable Wireless Ad Hoc Networks Routing Overheads are the non-data message packets whose roles are establishment and maintenance of routes for data packets as well as neighbourhood discovery and maintenance. They have to be broadcasted in the network either through flooding or other techniques that can ensure that a path exists before data packets can be sent to various destinations. They can be sent reactively or periodically to neighbours so as to keep nodes updated on their neighbourhoods. While we cannot do without these overhead packets, they occupy much of the limited wireless bandwidth available in wireless networks. In a reconfigurable wireless ad hoc network scenario, these packets have more negative effects, as links need to be confirmed more frequently than in traditional networks mainly because of the unpredictable behaviour of the ad hoc networks. We therefore need suitable algorithms that will manage these overheads so as to allow data packet to have more access to the wireless medium, save node energy for longer life of the network, increased efficiency, and scalability. Various protocols have been suggested in the research area. They mostly address routing overheads for suitability of particular protocols leading to lack of standardisation and inapplicability to other protocol classes. In this dissertation ways of ensuring that the routing overheads are kept low are investigated. The issue is addressed both at node and network levels with a common goal of improving efficiency and performance of ad hoc networks without dedicating ourselves to a particular class of routing protocol. At node level, a method hereby referred to as "link availability forecast", that minimises routing overheads used for maintenance of neighbourhood, is derived. The targeted packets are packets that are broadcasted periodically (e.g. hello messages). The basic idea in this method is collection of mobility parameters from the neighbours and predictions or forecasts of these parameters in future. Using these parameters in simple calculations helps in identifying link availabilities between nodes participating in maintenance of networks backbone. At the network level, various approaches have been suggested. The first approach is the cone flooding method that broadcasts route request messages through a predetermined cone shaped region. This region is determined through computation using last known mobility parameters of the destination. Another approach is what is hereby referred as "destination search reverse zone method". In this method, a node will keep routes to destinations for a long time and use these routes for tracing the destination. The destination will then initiate route search in a reverse manner, whereby the source selects the best route for next delivery. A modification to this method is for the source node to determine the zone of route search and define the boundaries within which the packet should be broadcasted. The later method has been used for simulation purposes. The protocol used for verification of the improvements offered by the schemes was the AODV. The link availability forecast scheme was implemented on the AODV and labelled AODV_LA while the network level implementation was labelled AODV_RO. A combination of the two schemes was labelled AODV_LARO.

Mobility-Based Routing Overhead Management in Reconfigurable Wireless Ad Hoc Networks Routing Overheads are the non-data message packets whose roles are establishment and maintenance of routes for data packets as well as neighbourhood discovery and maintenance. They have to be broadcasted in the network either through flooding or other techniques that can ensure that a path exists before data packets can be sent to various destinations. They can be sent reactively or periodically to neighbours so as to keep nodes updated on their neighbourhoods. While we cannot do without these overhead packets, they occupy much of the limited wireless bandwidth available in wireless networks. In a reconfigurable wireless ad hoc network scenario, these packets have more negative effects, as links need to be confirmed more frequently than in traditional networks mainly because of the unpredictable behaviour of the ad hoc networks. We therefore need suitable algorithms that will manage these overheads so as to allow data packet to have more access to the wireless medium, save node energy for longer life of the network, increased efficiency, and scalability. Various protocols have been suggested in the research area. They mostly address routing overheads for suitability of particular protocols leading to lack of standardisation and inapplicability to other protocol classes. In this dissertation ways of ensuring that the routing overheads are kept low are investigated. The issue is addressed both at node and network levels with a common goal of improving efficiency and performance of ad hoc networks without dedicating ourselves to a particular class of routing protocol. At node level, a method hereby referred to as "link availability forecast", that minimises routing overheads used for maintenance of neighbourhood, is derived. The targeted packets are packets that are broadcasted periodically (e.g. hello messages). The basic idea in this method is collection of mobility parameters from the neighbours and predictions or forecasts of these parameters in future. Using these parameters in simple calculations helps in identifying link availabilities between nodes participating in maintenance of networks backbone. At the network level, various approaches have been suggested. The first approach is the cone flooding method that broadcasts route request messages through a predetermined cone shaped region. This region is determined through computation using last known mobility parameters of the destination. Another approach is what is hereby referred as "destination search reverse zone method". In this method, a node will keep routes to destinations for a long time and use these routes for tracing the destination. The destination will then initiate route search in a reverse manner, whereby the source selects the best route for next delivery. A modification to this method is for the source node to determine the zone of route search and define the boundaries within which the packet should be broadcasted. The later method has been used for simulation purposes. The protocol used for verification of the improvements offered by the schemes was the AODV. The link availability forecast scheme was implemented on the AODV and labelled AODV_LA while the network level implementation was labelled AODV_RO. A combination of the two schemes was labelled AODV_LARO.