Analysis of P2P File-Sharing

The file diffusion process in P2P file-sharing networks can be compared to the spreading of a disease (e.g. avian influenza) in a finite population. Commonly, differential equations are used to describe the population dynamics, depending on the state the individuals are in. We use the well-known SIR model as basis for our model of how a file is disseminated in an eDonkey-like file-sharing network. Depending on the number of sharing users and the demand for a specific file, we can evaluate the rate of diffusion. Based on this mathematical model, our focus lies on the influence that malicious peers offering corrupt files have on the diffusion process (pollution).

Cooperative Content Distribution Networks in Cellular Networks

We evaluate the performance of a P2P-based content distribution system in heterogeneous, wireless networks. The mobile users coordinate each other with cooperation strategies enabled by the multi-source download mechanism, as in eDonkey or BitTorrent. Due to the mobility, vertical handovers between different wireless access technologies are required which may result in transmission delays and IP address changes of the switching peer. Hence, connections among users have to be reestablished and downloading users are requeued at a providing peer's waiting queue. In detail, we investigate a) the impact of requeueing with each VHO as well as the use of mechanisms that preserve the IP address and connections beyond VHOs, like MobileIP; b) the abrupt change of the available bandwidth, e.g., from a fast WLAN connection to a rather slow UMTS connection; c) the impact of user behavior, i.e. seeders and leechers. As a result, we develop different cooperation strategies to increase the efficiency and the robustness of the system while minimizing signaling traffic and guaranteeing fairness among users.

Supporting Mobility by P2P

Vertical handovers are expected to be a key feature in Beyond 3G (B3G) networks.We investigate the application of structured P2P overlay networks for supporting vertical handover in B3G networks. The P2P overlay is used to quickly locate attachment points for mobile entities and to retrieve rapidly the configuration and coverage information of these APs. The advantage of the P2P-based solution is its distributed nature, its scalability, and its self-organizing capability. In addition, this approach is applied for the configuration and management of radio access nodes in Beyond 3G (B3G) networks We compare the efficiency and the scalability of Pastry, CAN, and Kademlia and modify the protocols to optimize them w.r.t. these particular tasks.

Traffic Characteristics of Overlay Applications

The end-to-end quality of the communication between two end hosts can be evaluated on different levels and from different points of view. The traditional approach captures the QoS using measurements on the network layer. The derived technical parameters precisely describe the current ability of the network to provide a service but do not necessarily reflect the quality felt by the user of the service. On that account a new paradigm emerged which intends to asses the QoE describing the satisfaction of a user with the service. We give a more detailed description of the terms QoS and QoE with regard to different applications, show how they are related to each other, and how they can be measured in our testbed environment.The main focus of our studies is on how the current network conditions influence the QoE of the end user and in how far the application react to quality degradations. In this context, the network entity located in the middle of our testbed is used to emulate typical network environments, e.g. problems arising in wireless systems.

For example, we investigate Skype VoIP and some of the popular existing IPTV solutions by describing their architectures and characterizing their traffic through measurements. Among others, we focus on the applications Zattoo, Joost, YouTube, PPLive and OTR, and point out their differences in system architecture and technology. Especially in the case of proprietary systems, measurements at the edge of the network are the only feasible way to evaluate the QoS and QoE of an arbitrary user. Nevertheless, it allows us to identify the used protocols and constructed overlay topologies. E.g. Joost and PPLive use UDP for delivering the video content with a measured UDP downlink speed over 500 kbps in both cases. In contrast, Zattoo uses TCP for delivering the video content, also requiring over 500 kbps on the downlink. While Joost and Zattoo consume an upload bandwidth between 70-100 kbps, PPLive is the most aggressive in bandwidth demand (about 200 kbps) and number of contacted peers. While Zattoo only contacts about 100 peers and Joost about 500, PPLive connects to several thousands.The geographic locations of the contacted peers are locally confined in Zattoo due to licensing restrictions, but Joost and PPLive establish world-wide connections. In PPLive most peers are located in China due to the nature of the offered content. PPLive is the only application which offers both, live TV and VoD, while Zattoo provides only live TV and Joost only VoD.