Service Replication in Wireless Mobile Ad Hoc Networks

Mobile Ad hoc Networks (MANETs) are self configurable networks. A set of mobile nodes equipped with wireless communication capabilities can form a MANET. If two mobile nodes are located in the transmission range of each other, a wireless connection or link is formed. The set of the formed wireless links form the network topology. Since the mobile nodes are free to move, join and disjoin the network, the network topology is ever-changing. MANETs do not rely on any centralized administration or help to operate. Recently, MANETs have been gaining importance. The applications of MANETs are numerous and various. Personal area networks, Sensor networks and satellite networks are examples of this type of networks. The fact that MANETs do not rely on preexisting infrastructure may make them the only means of communications in some situations like search and rescue, disaster recovery and military scenarios. The mobile nodes work in cooperative and collaborative manners to achieve the required core network functionalities like routing and data transportation. The available resources of a MANET like the offered bandwidth between nodes, the computation power of the mobile nodes, and the power resources like batteries are typically very limited and always varying. Sharing resources in this type of networkis necessary. Service Oriented Architectures (SOAs) present a paradigm for achieving the required resource sharing in MANETs. If the resources to be shared are presented as service, the interested mobile nodes can access these resources through service requests. Recently, applying SOAs in MANETs has obtained a growing interest in research. SOAs offer many important features like the abstraction, encapsulation, de-coupling, discoverability, and autonomy which make them suitable for MANETs. In general, the great challenge for deploying services in a MANET is the service availability. Service availability is negatively affected by the previously mentioned features of MANETs. A service provider may not be reachable by some or all of the interested clients for some time as the network becomes partitioned because of the mobility of the nodes. The service provider can have a lack of resources to serve more client requests. 

In MANETs, some service providers can loose their power rapidly because of the high client interest of their services. These exhausted service providers may be quickly not available for the interested clients. Moreover, the scalability of the network affects the service availability. The paths between a service provider and a client are formed of many wireless links or \hops". The paths with higher number of hops are more likely to fail. A service provider mayappear unavailable to some clients because the paths to these clients are too long. Increasing the availability of the deployed services in MANETs can ensure a better overall performance for different applications. Generally, in distributed systems, one way to achieve this increased availability is through replication of services. Most of the service replication approaches evaluate their replication decisions based on performing analysis and predictions for the information of the network status. This required analysis is usually based on extracting some information from the lower network layers. Performing these computations is exhaustive and requires presence of special components like routing protocols or General Positioning System (GPS) antennas. In these cases, the design of the replication approaches or protocols is not only complicated but also architecture dependent. The main contribution of this thesis is therefore an architecture-independent service replication protocol for MANETs, Service Distribution Protocol (SDP), which relies on application-layer information, only. Throughout the thesis, this protocol and some enhancements are introduced and evaluated. Furthermore, this thesis introduces a set of formal definitions for the service replicability and replication process in MANETs. As an important contribution, the service popularity concepts and quantification methods are introduced. Also, measuring methods for the optimality of the service distribution in the network are proposed. The effects of the restricted resources from both client and provider sides on the proposed protocolare investigated and solutions are demonstrated and evaluated. Moreover, the proposed protocol has been involved to support the composite service execution process as a complex challenge for the availability of not only one service but also a set of services with different styles of execution flows.