Research Of Media Distribution Systems With Heterogeneous Peers

During the past decade, Internet technologies and personal computers have made great progress in terms of capabilities as well as costs. Ordinary online users not always trade emails and browse Web pages. They are increasingly often accessing a large variety of multimedia resources, for instance downloading audio/video files, watching live broadcasting, and demanding educational programs. Accessing multimedia demands much more bandwidth than that for emails and Web pages. In addition, the rapidly growing number of users makes the traditional client/server paradigm overwhelmed with respect to the upload bandwidth on the server side.As a resort, Peer-to-Peer (P2P) paradigm that is a fundamental principal of the Internet has been refocusing on. Compared to the traditional client/server paradigm, P2P organizes individual hosts at the edge of the Internet into overlay networks, and thus fosters an opportunity to harness various resources such as upload bandwidth, storage space, and computing cycles of these participating hosts (also called peers). As a result, system scalability is easily guaranteed.The main focus of this paper is on the investigation of how peer heterogeneity can influence media distribution systems in the P2P paradigm. We analyze the heterogeneity in detail and then propose countermeasures as well as algorithms, aiming at improving the performance and resilience of media distribution systems.The heterogeneity of peers in the overlay networks is well known. It can be described mainly by the physical capabilities and participating behavior. The physical capabilities of peers for instance upload bandwidth, storage space and computing cycles, differ significantly. Also, the connections among geographically different peers exhibit varied time delay and end-to-end bandwidth. As for the participating behavior, peers are highly autonomous and can thus make decisions independently. That means overlay networks almost certainly will face selfish behavior and malicious behavior, as well as obedient and altruistic behavior. Neglecting such behavior difference or not treating it properly will lead to performance degradation and, in some cases, make obedient peers being exploited by misbehaved peers.This dissertation studies the three typical scenarios of the media distribution systems, i.e., commons-based content distribution, on-demand media streaming, and live media streaming. It advances the state of the art in media distribution systems with heterogeneous peers. We in turn analyze how the leechers and seeds using the BitTorrent protocol are exploited by the selfish behavior and the malicious behavior, the weakness of on-demand media streaming services and streaming clients, and chunk scheduling strategies in data-driven live media streaming applications. Based on our analysis, we present the countermeasures and algorithmsThe main academic contribution and creativity of this paper is as follows:1. Using the symmetric encryption technique, a quota-based encrypted chunk trading protocol is proposed, which considerably discourages the free riding behavior, the most typical one of the selfish behavior. The new protocol can also resist the malicious behavior (e.g. offering fake chunks) to a certain extent. Furthermore, it does not rely on any third-party facilities, and completely conforms to the distributed characteristic of the BitTorrent protocol, facilitating its implementation and integration.2. Based on the combination theory and the symmetric encryption technique, a batch mechanism is presented to protect the seeds in BitTorrent. In this batch mechanism, the leechers can no longer acquire original chunks for free. Instead, they have to exchange (coded) chunks among themselves in the same batch (session) before being qualified to receive the necessary decoding information. It is shown that the batch mechanism can resist the free-riding behavior, and can also reduce the chance of attacks such as bandwidth-based ones and connection-based ones. Compared to the most related work, the batch mechanism has achieved an even higher level of robustness.3. Inspired by social-bookmarking techniques, popular on-demand/live media streaming applications are analyzed, highlighting their weaknesses as well as limitations of streaming clients. To overcome these issues, a general content service that is based on existing media streaming services is introduced. Being content-oriented in essence, this general content service can offer significantly more user-friendly viewing experience than existing services. Finally, promising extensions of the existing streaming clients are discussed.4. Upon identifying a drawback of an existing chunk scheduling algorithm in data-driven mesh-based live streaming applications, a new time constraint from among the set of blocks that can be requested is presented. After that, an exact formalization for the modified chunk scheduling algorithm is given in detail. At last, simulation experiments are conducted to validate the new algorithm.