Video Streams in a Computing Grid

Abstract: The growth of online video services such as YouTube enabled a new broadcasting medium for video. Similarly, consumer television is moving from analog to digital distribution of video content. Being able to manipulate the video stream by integrating a video or image overlay while streaming could enable a personalized video stream for each viewer. This master thesis explores the digital video domain to understand how streaming video can be efficiently modified, and designs and implements a prototype system for distributed video modification and streaming. This thesis starts by examining standards and protocols related to video coding, formats and network distribution. To support multiple concurrent video streams to users, a distributed data and compute grid is used to create a scalable system for video streaming. Several (commercial) products are examined to find that GigaSpaces provides the optimal features for implementing the prototype. Furthermore third party libraries like libavcodec by FFMPEG and JBoss Netty are selected for respectively video coding and network streaming. The prototype design is then formulated including the design choices, the functionality in terms of user stories, the components that will make up the system and the flow of events in the system. Finally, the implementation is described followed by an evaluation of the fault tolerance, throughput, scalability and configuration. The evaluation shows that the prototype is fault tolerant and its throughput scales bothvertically and horizontally. Intended audience This thesis focuses on topics in the area of general computer science and network technology. It is therefore assumed that the reader has knowledge of basic concepts and techniques in these areas. More specifically this report focuses on topics related to digital video and distributed computer systems. Knowledge in these areas is helpful but not required.