寬頻網路的普及與MPEG-4壓縮技術的發展促使網際網路上的資訊趨向更多元化的發展，而隨選視訊（Video-on-Demand）便是其中最熱門也最被看好的一項服務。然而，要將現有的影音格式轉換成MPEG-4規格需要非常長的壓縮時間，另外，雖然MPEG-4提高了壓縮比，隨選視訊系統仍需大量的儲存設備來存放影音資料，且目前與MPEG-4相關的硬體設備的價格仍然居高不下。而個人電腦叢集（PC Cluster）系統的架構可同時解決這三種問題。
本論文利用電腦叢集設計了一套隨選視訊系統，裡頭包含有串流伺服器與平行壓縮程式，並使用高普及率的MP4媒體規格及QuickTime Player當作影片格式與客戶端播放軟體。在MPEG-4壓縮方面該系統設計了一套低成本高效能提升率的平行壓縮架構，在高資料傳輸量之下使用平行輸出/輸入（Parallel I/O）降低整個平行計算過程中各主機等待接收資料的時間進而得到高效能的平行壓縮架構，並透過實驗證明該架構確實擁有較好的效能提升及討論其優缺點。
在串流伺服器方面該系統採用二階層（Two-Tier）架構，並利用叢集系統高度可擴充性（Scalability）的特性作為串流伺服器中的儲存系統（Storage System），使用資料分散（Data Stripping）的媒體資料儲存方式分散磁碟負載，並討論在不同的硬體環境下該如何建置與擴充該串流伺服器。最後我們利用MySQL做為媒體資訊資料庫並提供一個Web環境的使用者界面與管理界面讓這個系統更容易被使用與管理。

With the wide deployment of wide-bandwidth network and the advance of the MPEG-4 compression technique, many types of the information service in WWW are developed and provided. Video-on-demand is the hotest and the most potential one among these services. However, it costs a long time to compress a normal video file to a MPEG-4 file. In addition, although the MPEG-4 has the high compression rate, video-on-demand system still needs a great amount of the storage device for storing the video data, and the price of this device is expensive. Fortunately, using PC Cluster to build video-on-demand system can eliminate these issues.
In this thesis, we develop and design a video-on-demand system based on the PC Cluster. The system includes streaming server and parallel-compressing process. Furthermore, the system adopts a popular MP4 media format and QuickTime Player for the movie format and the player software in the client respectively. We develop a parallel compression technique which has high performance but relatively low cost in the system. The proposed parallel input/output architecture is used to reduce the waiting time of receiving data to the processing of parallel computing for every computer under the high rate of data transforming. That is, we have obtained parallel compression architecture with high performance. From the experimental results, it is indicated that the performance of the parallel compression architecture is better than that with serial input/output parallel compression architectures. Simultaneously, the advantages and defects of the proposed architecture are also shown from the experimental results.
We adopt a two-tier architecture to construct the streaming server in the proposed system. We use the PC Cluster to be the storage system of the streaming server because it’s scalability characteristic, sharing disk loading by data stripping storage method. We also discuss how to construct and scale this storage system under different hardware environment. Finally, we adopt MySQL database and provide a Web interface for clients and the administrator for the facility and management of the system.

[1] Chao-Tung Yang, Chia-Chen Soong, Shian-Shyong Tseng, “A Study of Multithreaded Programming on Linux SMP Architecture”, The Sixth Workshop on Compiler Techniques for High-Performance Computing,March 16-17, 2000
[2] Chao-Tung Yang, Chia-Chen Soong, Shian-Shyong Tseng, “A Linux/PVM Cluster on Low Cost SMPs Architecture”, The Sixth Workshop on Compiler Techniques for High-Performance Computing, March 16-17, 2000
[3] Christoph Bernhardt, Ernst Bierscak, “THE SERVER ARRAY: A SCALABLE VIDEO SERVER ARCHITECTURE”, Chapter 5 in “High-Speed Networking for Multimedia Application”, O. Spaniol, W. Effelsberg, A. Danthine, D. Ferrari, eds., Kluwer, Publ., March 1996
[4] Dapeng Wu, Yiwei Thomas Hou, Member, Wenwu Zhu, Member, Ya-Qin Zhang, and Jon M. Peha, “Streaming Video over the Internet: Approaches and Directions”, IEEE Transactions on Circuits and System for Video Technology, VOL. 11, No. 3, March 2001
[5] “Grid Computing”, http://www.globus.org
[6] Hai Jin, Guang Tan and Song WU, “Clustered Multimedia Servers: Architecture and Storage System”, Internet and Cluster Computing Center, China, 2003
[7] ISO/IEC,”ISO Media File format specification-MP4 Technology under consideration for ISO/IEC 14496-1:2001/Amd 3”, JTC1/SC29/WG11 MPEG01/N4270-1, July 2001
[8] Miguel Ribeiro, Oliver Sinnen, Leonel Sousa, “MPEG-4 Natural Video Parallel Implemetation on a Cluster”, RECPAD 2002－12th edition of the Portuguese Conference on Pattern Recognition of APRP - The Portuguese Association for Pattern Recognition
[9] “MPICH- A Portable Implementation of MPI”, http://www.unix.mcs.anl.gov/mpi/mpich
[10] Peter S. Pacheco, “A User’s Guide to MPI”, Department of Mathematics University of San Francisco, March 1998

[11] Renu Tewari, Rajat Mukherjee, Daniel M. Dias, Harrick M. Vin, “Design amd Performance Tradeoffs in Clustered Video Servers”, International Conference on Multimedia Computing and Systems, 1996
[12] R. Buyya, “High Performance Cluster Computing: System and Architecture”, Vol. 1, Prentice Hall PTR, NJ, 1999
[13] R. Buyya, “High Performance Cluster Computing: Programming and Applications”, Vol. 2, Prentice Hall PTR, NJ, 1999
[14] RFC 2327 : “SDP: Session Description Protocol”
[15] RFC 2326 : “Real Time Streaming Protocol（RTSP）”
[16] RFC 1889 : “RTP: A Transport Protocol for Real-Time Application”
[17] Sugih Jamin, Scott J. Shenker, Peter B. Danzig, “Comparison of Measurement-based Admission Control Algorithms for Controlled-Load Service”, EECS Department University of Michigan Ann Arbor, 1997
[18] The QuickTime Technique Reference Library, “Inside QuickTime: QuickTime File Format”, http://www.quicktime.com , March 2001
[19] 趙勗辰，”以Sage++發展Web環境之C循序語言自動平行執行環境”，國立成功大學資訊工程研究所，民國90年
[20] 王宏瑾，”MPEG-4 伺服器與同步機制的設計與實作”，國立台灣大學電機工程研究所，民國88年