Techniques for streaming, storage and search of video.

Recent developments like better last-hop bandwidth to homes, free video sharing websites, availability of cheap consumer electronics to produce and consume video have led to wide proliferation of video in many aspects of our lives. This leads to many interesting technical challenges that forms the broad topic of this thesis. In this thesis, we make four key contributions towards providing video services over the Internet. (1) Bandwidth and delay properties of a shared, global network like the Internet constantly change over time. Providing Quality of Service (QoS) for Internet-based video streaming applications requires the server and/or client to be network-aware and adaptive. We propose a video server architecture that can adapt to changes in the network conditions with minimal effect on client's perceived video quality. (2) With an increasing variety of video consuming devices, each with its own strengths and limitations, content needs to be tailored to fit every device's capability. We propose a new agent called composing and transcoding (CaT) gateway that intelligently uses two types of video tailoring techniques to suit content for devices better. (3) Delivering video to millions of clients requires large scale bandwidth commitments from a server. With large number of users and phenomena like flash crowds, peer-to-peer (P2P) streaming is a very attractive option to provide scalable streaming services. However, with P2P streaming, scalability comes at the cost of reliability and quality of the delivered video. We propose a framework called StatStream that provides statistical guarantees on the reliability and quality of video delivered through a P2P overlay network. (4) Video sharing websites require good search methods to browse for videos of interest. In that context, richer query models like content-based search techniques look very promising. We point out why todays' systems are not well equipped to handle content-based searches. We propose a new system called SQUAD that can handle content-based searches better. SQUAD also provides a richer query interface to search video metadata in addition to the video content itself.