Cooperation Strategies for P2P Content Distribution in Cellular Mobile Networks: Considering Selfishness and Heterogeneity

Introduction

P2P file sharing systems contribute to the majority of traffic volume that is currently being transported in the Internet. Applications like eDonkey or BitTorrent are used to share large volume content and alleviate the problem of overloaded servers by distributing the load among all sharing peers, which makes P2P systems highly scalable and resilient content distribution systems. The performance of such P2P content distribution networks (CDN) in cellular networks depends highly on the coordination of heterogeneous and often selfish mobile users. Sophisticated cooperation strategies, such as the multi-source download and tit-for-tat principle, are the foundation of the extreme efficiency of P2P content distribution networks. Multi-source download means the simultaneous download of parts of a file, referred to as chunk, from several sources in parallel. The cooperation strategies applied in popular P2P CDN platforms such as eDonkey or BitTorrent, rely on the fundamental P2P assumption that all peers are equal. In cellular networks, however, the peers differ significantly in their characteristics, e.g. their access system and bandwidth which might change over time or their on-line behavior, thus introducing heterogeneity and even selfishness in the peer community. Hence, the P2P assumption of equal peers is not valid any more. In addition, the dynamics and heterogeneity in cellular mobile networks is further increased by the mobility of users.

Although most P2P CDNs utilize the benefits of multi-source downloads, the various platforms differ significantly in the actual implementation of the cooperation algorithms. In particular, the peer selection as well as the chunk selection mechanisms lead to different system behaviors and performance results. The detailed performance of the strategies is further determined by the actual peer characteristics and the peer behavior. The peer characteristic includes, among others the available upload and download bandwidth, as well as the number of parallel upload and download connections. The mobility of a user makes these peer characteristics change over time. Thus, the performance depends considerably on the heterogeneity. The peer behavior is mainly described by churn, i.e. the switching of a user between offline and online state, and by the willingness of a user to participate in the CDN. A user may behave selfishly and tries to minimize the upload of data or he may redistribute the data in an altruistic way. In the context of cellular mobile networks, churn and selfish behavior appear even more distinctive, e.g. to save battery resources or scarce and expensive uplink capacities. As a result, the so-called “last chunk” problem might arise which inhibits the data dissemination process and makes individual chunks starve in the network.

Additional challenges and influence factors on the performance of the system arise in a heterogeneous, wireless cellular network. Due to the user mobility, vertical handovers (VHO) between the different wireless access technologies are required which may result in transmission delays and IP address changes of the switching peer. Mechanisms like Mobile IP allow overcoming IP address changes, nevertheless such mechanisms also introduce additional delays. Another important phenomenon occurring with vertical handovers is the abrupt change of available bandwidth, e.g., from a fast WLAN connection to a rather slow UMTS connection. This will be discussed in more detail in the coherent following chapter “Cooperation Strategies for P2P Content Distribution in Cellular Mobile Networks: Considering Mobility and Heterogeneity”.

There are several possibilities to improve the performance of content distribution in cellular networks. Those are: a) particular architecture concepts introducing special entities like caches for storing contents or crawlers for locating sources, e.g. (Oberender et al., 2005); b) the optimization of parameters, like the size of chunks, as done by Hoßfeld, Tutschku & Schlosser, 2005); c) incentives to motivate the users to share files and to contribute to the system; and d) cooperation strategies for the coordination among peers. From these possibilities, we will focus on the cooperation strategies in this chapter. The goal is: i) to describe how to model a P2P content distribution system with multi-source download in a cellular wireless environment; ii) to identify the fundamental problems of typical cooperation strategies; iii) to investigate the impact of user behavior and heterogeneity, in particular selfishness; and iv) to propose solutions to overcome the derived problems.