Design and Implementation of a Video Transcoding System in Cloud Computing

• Main Authors: Jun-yi Wu, 吳俊逸
• Other Authors: Jiann-Jone Chen
• Format: Others
• Language: zh-TW
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/72691179280922971051

碩士 === 國立臺灣科技大學 === 電機工程系 === 102 === Personal mobile communication devices become popular in recent years. Smartphones and tablet PCs are used to browse webpages or watch live streaming video through the Internet, such as the games live, distance education, etc. How to achieve a compromised perception quality under this condition is a challenging issue. However, the transmission is unstable under wireless/mobile network environment, and the user devices may not support high quality video consumption. To solve this problem, it needs to transform the video format from one to another that is supported by the target devices, i.e., video transcoding. As video transcoding is a computationally intensive process, it is usually carried out based on a cloud computing platform to achieve fast transcoding for more than one users with different end devices. A MPEG-DASH (Dynamic Adaptive Streaming over HTTP) standard, which can dynamically adapt media streaming over HTTP, had been proposed by MPEG to make video streaming compatible with different end devices. It has been widely used in various video streaming application systems, such as Youtube, YouKu and VideoLAN etc. In this research, a cloud transcoding system is designed that support MPEG-DASH to serve users with different devices and bandwidth environment. One input video can be transcoded to another format with different resolution, bitrates and frame rates to meet user’s requirement. For cloud computation, the input video would be divided into several smaller video segments, and we investigate how to perform the division to yield the minimum system transcoding time. In addition, we proposed a Multi-layer Division of Max-MCT scheduling algorithm (MDMCT) to reduce the overall job’s convoy effect. It helps to coordinate the system operations and dynamically adjust the number of slots so that the cloud clusters can finish the process more efficiently. Experimental results show that the proposed method can help to achieve better system load balancing. The entire cloud clusters of resource utilization maintained at about 98%, and can be shortened by about 10%-13.4% of the transcoding time.