Peer-to-peer overlay topology control for mobile ad hoc networks

Abstract

The enormous popularity of peer-to-peer (P2P) applications and the increased use of mobile devices make running P2P applications on mobile ad hoc networks (MANETs) natural. However, simply applying existing P2P overlay techniques, which are designed for wired networks, to MANETs is undesirable due to the differences between the two types of networks. In P2P systems, peers are often selfish rather than cooperative (which is expected of MANET nodes). In this paper, we study the construction of P2P overlays by selfish peers in the context of MANETs, and propose a computationally feasible game-theoretic heuristic algorithm. In our P2P-MANET creation game, peers seek to maintain as few neighbors as possible while minimizing their distances to all destinations, in an effort to reduce energy consumption and to improve response time. We find that Nash equilibria are difficult to find at best, and may not even exist in most cases. Our heuristic is fairly stable relative to the minimum cost algorithm, and when the degree–constraint is relaxed, it approximates the minimum cost. The lack of global knowledge of the overlay and underlying network at individual peers does not allow each peer to fully exploit such information for its own best interest, and actually may reduce the total cost network-wide.

Introduction

P2P applications are enormously popular on the Internet and their uses vary from file sharing to Voice-over-IP to gaming and more. Increasingly, users are moving from wired networks toward mobile networks. Mobile ad hoc networks (MANETs) are essential in some environments, such as in disaster recovery, or military battlefields, where the communications infrastructure has been damaged or is non-existent. In other environments, such as at a company meeting where the participants would like to share documents, such networks may prove more convenient than using a wireless LAN. Recently, the Wi–Fi Alliance has announced a new specification called “Wi–Fi Direct” that allows devices to connect to one another directly, without the use of a base station [1]. The code-name for the specification was “Wi–Fi peer-to-peer”. A natural evolution is for P2P applications to run on a cooperative MANET (we use the term P2P–MANET to refer to such networks).

Simply applying existing P2P overlay techniques to MANETs is undesirable due to the lack of infrastructure, node mobility, and energy issues. It is important that the overlay topology reflects the underlying network, both to reduce energy consumption and to improve response times. Nodes in MANETs are mobile, meaning that the underlying topology is constantly changing as users move about. Existing P2P topology control schemes, designed for wired networks, are able to accommodate a changing topology due to the expectation that peers will constantly be joining and quitting. However, nodes are assumed to be disconnected once they are unreachable in non-mobile networks. In MANETs, this assumption is incorrect since nodes may have simply moved to a different location. Mobile devices are powered by battery. The overlay control algorithm needs to be computationally simple and ensure that MANET nodes forward as less traffic as possible. On the one hand, it is preferable for peers to have a small degree on the overlay to reduce control traffic. On the other hand, it is preferable for the overlay to be well-connected so the traffic between two overlay peers traverses as few MANET nodes as possible. Because of the lack of infrastructure, a fully distributed overlay topology control algorithm is preferred.

Most P2P overlay topology control algorithms assume that peers are cooperative. Unfortunately, peers are selfish in many cases. They seek to minimize their own costs, in the case of MANETs, to minimize the number of links to other peers and the distance to all destinations. Several studies [2], [3], [4] investigate the impact of selfish peers on the topology in the context of non-mobile networks. However, they study the theoretical bounds or require peers to have global knowledge to construct the overlay. No practical overlay topology control algorithm, even for non-mobile networks, exists. As well, a study of the impact of selfish peers in the context MANETs is needed.

In this paper, we study the construction of the P2P overlay by selfish peers in the context of MANETs, and propose a fully decentralized, computationally feasible, heuristic algorithm based on game theory. In our P2P–MANET creation game, peers seek to maintain as few neighbors as possible while minimizing their distances to all destinations. We compare the topology of our heuristic to that of random local search algorithms. We find that Nash equilibria are difficult to find or may not even exist. Minimum cost topologies are typically highly connected and thus resilient, but are not in a Nash equilibrium state. Our heuristic is fairly stable relative to the minimum cost algorithm, and when the degree–constraint is relaxed, it approximates the minimum cost. The lack of global knowledge of the overlay and underlying network at individual peers does not allow each peer to fully exploit such information for its own best interest, and actually may reduce the total cost network-wide.

The remainder of the paper is organized as follows. Section 2 provides related work on P2P overlay topology control algorithms. Section 3 discusses the game-theoretic heuristic overlay topology control algorithms in detail. In Section 4 we compare the topologies formed by the heuristic and benchmark algorithms. Section 5 provides some conclusions.