THESIS
2008
xiii, 75 leaves : ill. ; 30 cm
Abstract
This thesis work proposes two novel techniques, practical random linear network coding based video data delivery scheme and bottleneck-free overlay network formation algorithm, to overcome the two fundamental problems in P2P VoD networks....[
Read more ]
This thesis work proposes two novel techniques, practical random linear network coding based video data delivery scheme and bottleneck-free overlay network formation algorithm, to overcome the two fundamental problems in P2P VoD networks.
The first problem in the P2P VoD networks is the data block availability problem which are caused by lacking of fresh data among connected neighbor peers. If there are no useful data in the neighbor peers, the uploading bandwidth of the neighbors can not be fully utilized. It has been shown that random linear network coding can efficiently solve this problem by replacing the data block scheduling and downloading mechanism with network coding data delivery scheme. However, due to the tremendous coding complexity, the random linear network coding codec is never fast enough to perform large scale data encoding and decoding. To solve this impracticality problem, we propose a novel lightweight random linear network coding video data delivery scheme which has a much lower coding complexity. In the proposed random linear network coding codec, the coding processes implement Galois field 2 operations which mainly consist of bit-wise binary operations. The processing speed is greatly increased. Moreover, we also propose to limit the number of original data blocks which are used in the encoding process to further reduce the coding complexity and the processing speed increases significantly. It is widely believed that network coding with a low degree Galois field has a much higher probability of generating dependent packets. However, we prove that the Galois field 2 (the lowest degree Galois field) random linear network coding has very limited number of dependent packets generated in the coding process. The contribution of this part of our work is that we have finally achieved a practical random linear network coding codec which has the ability to process large amount of data without consuming too much processing resource. The excellent coding speed is crucial to support smooth video streaming in P2P VoD networks.
The second problem of the P2P VoD networks is the unbalanced uploading bandwidth bottlenecks which are caused by insufficient uploading bandwidth of the overloaded neighbor peers. We propose a novel overlay network formation algorithm which constructs a perfect workload balanced overlay network among active peers. In the overlay network constructed by our algorithm, each uploading peer serves an appropriate number of child nodes so that the downloading connections in the overlay networks obtain the same level of bandwidth. As a result, the video data delivery does not encounter any bandwidth bottlenecks. Moreover, we take the attribute of asynchronized transmission in P2P VoD networks into consideration when we construct the overlay networks so that the video data can be forwarded from generation to generation without experiencing any blockage. By combining the previously proposed random linear network coding video data delivery scheme with the bottleneck-free overlay network formation algorithm, the downloading performance of peers is guaranteed. As a result, the near-optimal throughput for P2P VoD systems can be achieved.
Post a Comment