Investigating the limitations of video stream scheduling in the Internet

Abstract

Streaming media files over the Internet have become highly popular in recent years, and the requirements for network resources, and client and server hardware have grown. In this thesis, we argue that today’s unicast delivery scheme for video does not scale well when the number of clients is growing. The need for stream scheduling schemes is increasing as more media data is made available on the Internet and the number of clients using online media services is rising.

The main contributions in this thesis is to investigate a patching scheme in a multicast environment for stream scheduling in the Internet. We have implemented a server that provide the clients with video in several qualities over multicast channels. We have also implemented a client that is responsible for receiving and consuming the video from the multicast stream. It is also responsible for downloading patch streams from independent web servers for the parts of the video that are missing from the multicast stream. The client use quality adaptation to receive the video in best possible quality based on the client's network resources.

The Internet is not a scene without any challenges or problems. Challenges like packet loss, jitter and heterogeneous Internet connections amongst the clients are present. We have investigated how these challenges affect our implementation in a real network envi- ronment.

Based on the results from our experiments, we have concluded that our implementation of the patching scheme is able to cope with the challenges that we know exist in the Internet. By the use of a quality adaptation scheme, the user is able to receive video in the best quality possible with regard to these challenges.