| Summary: | 博士 === 國立清華大學 === 資訊工程學系 === 99 === Many researches on peer-to-peer video streaming have focused on dealing with highly dynamic, high-churn P2P environment. There have been tremendous efforts and innovations in mesh-based (also known as data-driven or pull-based) P2P streaming protocols and some of them were modified from a P2P file sharing protocol, called BitTorrent. Inspired by the high performance on peer-to-peer file sharing of the BitTorrent, we propose a scheme for overlaying streaming mechanisms on the native BitTorrent protocol. In the first part of dissertation, we focus on a practical P2P video streaming system-WuKong and investigate the stream delivery algorithms for each component. WuKong not only takes advantages of the BitTorrent but also combines the video scalability of layered video coding. An adaptive layer-downloading process is investigated to balance between the video quality and bandwidth utilization on heterogeneous peers. WuKong is implemented on using an open-sourced library of the BitTorrent protocol and coding schemes of the Windows Media Video(WMV) and the Scalable Video Coding(SVC). We measured and compared the end-user quality of WuKong on heterogeneous peers in this system. In addition, we evaluated the effectiveness of WuKong with randomly peers that are joining and leaving. The results show that WuKong not only provides high quality P2P video streaming services but also supports different scaling abilities over heterogeneous devices.
In the second part of the dissertation, we focus on investigate a high-availability BitTorrent tracker. Entering and leaving of random peers will result in great variance of data availability in P2P networks, especially when multimedia streams require high priority to the immediate playback data. All peers must be able to receive the complete video segment before that playback. Therefore, the provision of the tracker in the P2P networks is to accurately track the status of each peer and so to maintain high data availability being considered functionally. Our work deliberates the redundant mechanism of BitTorrent tracker and designs a BT-like redundant tracker mechanism that is configured to this purpose, and presents the relative performance measure. The result shows that the proposed redundant tracker does not increase overheads and is able to upgrade stability to the playback operations in P2P streaming networks.
The capacity of P2P streaming systems, \textit{i.e.}, how many peers can be concurrently serviced by the system, depends on system configurations including network bandwidth, overlay network formation, QoS constraints, membership management, peer selection, subscribing/publishing scheduling, coding schemes, etc. Due to the complexity, there are not too many systematic studies on system capacity. The third part of the dissertation answers this question of our system. Based on flow concepts, we compare the capacity of single-tree P2P streaming systems and multiple-tree P2P streaming systems. We prove that finding the routing tree to maximize system capacity, no matter for either single-tree systems or multiple-tree systems, is NP-hard. Two generic network formation heuristics, one for single-tree topology and the other for multiple-tree topology, are proposed. Simulation results show that since multiple-tree topology can efficiently utilize upload bandwidth, the multiple-tree systems have potential to service more clients. Moreover, we exposed the bandwidth and delay constraint on achieving maximum system capacity. The centralized and distributed network formation heuristics were proposed. We compared the performance of our centralized and distributed heuristics with the maximum uplink/minimum delay first scheme. Simulation results showed that our algorithm can achieve higher system capacity, since the algorithms considered on both uplink capability and delay constraints.
|
|---|
