DEBUT: Delay bounded service discovery in urban Vehicular Ad-Hoc Networks

This paper studies delay-bounded service discovery in urban Vehicular Ad-hoc Networks (VANETs), which refers to locating resources and services (e.g., local sensor data and multimedia content) distributed on individual vehicles in the network within a certain delay bound. To facilitate the discovery process, a set of vehicles, called service directories (SDs), can be selected to store the index information of all the resources in the network. Selecting an optimal SD set with minimal size while satisfying the users' requirement of a bounded query response delay is very difficult due to the disruptive nature of VANETs. In this paper, we formulate the Delay Bounded Service Directory Selection (DB-Sel) problem as an optimization problem that minimizes the number of SDs under the delay bound constraint. We prove theoretically that the DB-Sel problem is NP-Complete even when the future positions of vehicles are known a priori. We observe and prove that the number of vehicles encountered by arbitrarily selected SDs within a given delay follows a normal distribution. We also find the contact probabilities among the vehicles exhibit strong temporal correlation. With these observations, we develop a heuristic algorithm which iteratively selects the best candidate according to the normal distribution property and the historical contact probability. We prove that our algorithms have a guaranteed performance approximation ratio compared to the optimal solution. Extensive trace-driven simulation results demonstrate that our algorithm can guarantee the required query delay and select SD sets 20% smaller than those selected by alternative algorithms.