Back to results list
Show full item record
Please use this identifier to cite or link to this item:
|Title:||Providing robust and cost-effective large-scale video-on-demand services in networks||Authors:||Sze, Kwok Tung||Degree:||Ph.D.||Issue Date:||2017||Abstract:||Video-on-Demand (VoD) service allows users to view any video provided by the service provider at any time and to enjoy the flexible control of video playback. However, the deployment of a large-scale VoD system demands an enormous amount of resources. One of the most challenging design aspects is how to deliver videos to a large community in a cost-effective manner. This thesis presents the results of our works in this area in order to build a robust and cost-effective VoD systems. A feasibility study on supporting the VoD service over Wireless Mesh Network (WMN) is first carried out. In the proposed system, a video is divided into blocks, and the copies of those blocks are replicated and distributed to the routers in the network evenly. The required copies of each video block decrease with its timing order, i.e., fewer copies are needed for video blocks in higher order. Each router only stores a portion of the video. A video discovery scheme is developed to search the video blocks from the network. Different transmission strategies are considered that introduce three possible enhancements, named Video Block Broadcasting, Peer-to-Peer, and Time-Shifting. These enhancements aim to utilise the bandwidth of the routers so that more users can be served concurrently. A simulation model is built to study the performance gain of each enhancement regarding blocking probability. Simulation results show that the proposed enhancements can achieve different levels of improvement. The Peer-to-Peer and Time-Shifting enhancements are suitable for all situations while the Video Block Broadcasting can only be used when the VoD system already has a substantial number of free channels. As a result, this motivates us to further study the potential on combining these enhancements. A PB-P2P VoD system is then proposed. The system is a hybrid of Periodic Broadcasting (PB) and Peer-to-Peer (P2P) mesh pull. The proposed system aims at minimizing the startup delay, maintaining smooth playing, and supporting various interactive commands at best effort. The resources demanded by PB are independent of the number of incoming requests, and the service capability of P2P is increasing with the number of servicing users. Therefore, both techniques have good scalability that is suitable for delivering the popular videos to a large community.
In the system, a video is divided into segments which may be different in length according to the segmentation scheme in use. Each segment is then divided into a number of chunks with equal length. Server uploads the chunks to channels periodically. Clients switch among these channels and query peers in the same network concurrently to download the chunks. Scheduling algorithms are developed for clients to coordinate downloads from both channels and peers simultaneously and seamlessly under various conditions. The system is implemented in ns-3. Simulations are conducted to examine the impact of different factors to the system performance. Simulation results show PB and P2P are a good complement to each other. P2P is critical to shorten the startup delay, improve the execution of the interactive commands, and recover the corrupted chunk. These characteristics complement the weaknesses of the PB. PB guarantees the maximum of startup delay, ensures smooth playback if no interactive command is involved, and provides an efficient and stable source of video data. These characteristics are difficult to achieve by using P2P alone. The final work of this thesis is to study the potentials on using the multicast in the P2P network. Although the PB-P2P technique could build a robust and cost-effective VoD system, no every user would contribute his resources such as upload bandwidth to the VoD system in practice due to various reasons. Moreover, the P2P servers upload the chunks through unicast that can only serve a limited number of clients at a time. The multicast could be a potential solution that allows a P2P server to accept and satisfy requests from different clients for the same chunk by single upload stream. Simulation results show that Peer-to-Peer over Multicast (P2P-M) could achieve better performance when the user contribution rate is low and the number of P2P server is limited.
|Subjects:||Hong Kong Polytechnic University -- Dissertations
Wireless communication systems
|Pages:||xxvi, 157 pages : color illustrations|
|Appears in Collections:||Thesis|
View full-text via https://theses.lib.polyu.edu.hk/handle/200/9031
Citations as of May 22, 2022
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.