Data caching in ad hoc networks using game-theoretic analysis

Abstract

There have been researches that studied selfish data caching in ad hoc networks using game-theoretic analysis. However, due to the caching problem’s theoretical root in classic facility location problem and k-median problem, most of the researches assume: 1) The data is initially outside of the network; 2) The caching cost is either a constant or not considered at all. In reality, there are many applications, such as ad-hoc and sensor networks and peer to peer networks, in which data is initially collected or stored in the network and the caching cost depends on the network topology. This thesis addresses the problem of in-network data caching (referred to as in-caching problem) in multi-hop stationary ad hoc networks where the data is initially stored in the network and both caching and accessing costs are distance dependent. We first show that the problem is NP-hard. For selfish data caching game of the problem, we show that a pure Nash Equilibrium exists, in which a node will not deviate its caching strategy if others keep their own strategy. However, a Nash Equilibrium may not guarantee social optimal cost – due to the selfishness of each node, the price anarchy, which is the relative cost of the lack of cooperation among nodes, could be as large as O(n), where n is a number of nodes in the network. Using an external incentive mechanism based upon a payment model, we show a Nash Equilibrium and social optimal can both be achieved simultaneously via extensive simulations.