| Summary: | 碩士 === 國立中正大學 === 資訊工程學系 === 85 === In a video on demand system, video data can be transmitted based
on CBR or VBR. If CBR is used, a large buffer is required at the
client. Hence, most of VOD systems adopt variable bit rate to
transfer video program. However, traffic parameters of VBR is
difficult to characterize and QoS is hard to predict. In this
thesis, we have proposed a partition method that can partition a
video program into minimum number of segments while each segment
is transmitted by CBR. Base on this method, the buffer space at
client can be reduced and the overhead of re-negotiation of
bandwidth can be decreased because the number of video segments
is minimized. Although this method can partition a video program
into minimumnumber of segments, the number of segments may still
very large, due to the small size of client's buffer. To
overcome this, we have also proposed a merge method in this
paper. This method can merge several adjacent segments into one
pseudo segment such that the number of rate changes is reduced.
Furthermore, the proposed method not only reduces the number of
rate changes but also minimizes the wasted bandwidth. Finally,
we also proposed a method for distributing all video programs to
different level of servers in a hierarchical VOD system while
minimizing the operation cost.
|
|---|
