Modeling and simulation of service composition in opportunistic networks

Pervasive networks formed by users' mobile devices have the potential to exploit a rich set of distributed service components that can be composed to provide each user with a multitude of application level services. However, mobile and pervasive networks suffer from intermittent connectivity, disconnections and partitions, such that opportunistic networking techniques are required to enable communication. This poses novel challenges to service composition techniques. While several works have discussed middleware and architecture for service composition in well-connected wired networks and in stable MANET environments, the underlying mechanism for selecting and forwarding service requests in the significantly challenging networking environment of opportunistic networks has not been addressed. The problem comprises three stages: i) selecting an appropriate service sequence set out of available services; ii) forwarding service inputs to the device hosting the next service in the composition; and iii) routing final service outcomes back to the requester. The proposed algorithm derives efficiency and effectiveness by taking into account the service load and location of devices providing the services, as well as intermittent connectivity, to select a particular service set. Through extensive simulations on real and synthetic traces, we show that by using only local knowledge collected in a distributed manner, performance close to a real-time centralized system can be achieved.