現有的隨選視訊技術,都是採用專屬串流的方式來服務單一使用者,如此一來,不但造成頻寬的浪費,當多個使用者同時提出相同的服務需求時,有可能會因為伺服器端的頻寬不足而造成使用者的過度等待,進而減低使用者觀賞的意願。傳統的資料廣播機制,主要是先將影片分成許多片段(segment),同時在多個資料頻道上週期性的廣播,其中一個頻道會即時傳送第一個片段,而其他頻道則傳送剩下的片段,本論文提出了以反向序列廣播法建構隨選視訊伺服器之研究與實作。依據我們所設計的廣播演算法Reverse-order Scheduling (ROS) method,提供一低等待時間、低緩衝需求以及低成本的廣播系統,並實際應用在現有的IPTV架構中,提供使用者熱門影片的廣播服務,使用者可以透過終端裝置,如電腦、Set top box等取得隨選視訊的服務。ROS演算法第一步是將影片依照系統的頻寬切割成多個相等長度的影片片段。第二步是將頻道切割成多個groups,並切割好的影片片段依反向的順序在這些subchannel中廣播。從實驗的結果分析,可以將使用者的緩衝需求降至36%,當頻寬使用3b以上(b is consumption rate),並用300 Second (s)、600(s)、900(s)的片段進行實驗,其平均等待時間24 (s)、47(s)、70(s),在最大緩衝需求28%、36%、29%,多使用者分別測試2人、3人、4人、5人實驗結果的等待時間,並證實系統的可行性。
The existing video-on-demand (VOD) techniques usually dedicate a bandwidth to a single user, and this is not only using the bandwidth inefficiently but also increasing the waiting time when multiple users are simultaneously raising the same demand for the service and the bandwidth would become insufficient. And with a longer waiting time, the users would be less motivated to watch the video. Therefore, how to utilize the limited bandwidth to serve more users has always been the subject of academic and industry studies. Broadcasting has always been one of the most effective techniques for VOD applications. Traditional broadcasting systems broadcast segments of a video periodically on different channels simultaneously. In this paper, we have developed a broadcasting based VOD on a single channel with low waiting time and low buffer requirement. This system will utilize the movie's popularity to determine how much bandwidth is needed to broadcast the video, designed according to our broadcasting algorithm “Reverse-order Scheduling (ROS)” method. It provides a low latency, low buffer requirements, and low cost system, and a practical application for the existing IPTV architecture, which provides users with popular video broadcasting services. With our system, users can obtain video on demand and other services with terminal units, such as computers and set top box, without worrying about a very long waiting time. The ROS method firstly partitions the video into equal-length segments, then divides these segments into groups, and finally broadcasts the segments in the same group over the same sub-channel in the reverse order of their indices. In accordance with our experiment analysis, we have been able to reduce the user's buffer requirements to 36%. When the bandwidth is up to 3b (b for consumption rate), we use 300(sec), 600(sec), and 900(sec) length videos for experiment, and get the average waiting time of 24(sec), 47(sec), and 70(sec), while the maximum buffer size requirement at 28%, 36%, and 29%. We also tested the waiting time in multi-users environment with 2 to 5 clients connecting to the server, and proved the feasibility of the system.