This thesis presents the design and implementation of a novel technique in providing quality of service in transferring multimedia data over the Internet. With the advent of IP multicast and increased use of real-time multimedia applications on the Internet which rule out the use of Transmission Control Protocol (TCP), high packet loss rates have been observed in non-TCP protocols. A number of forward error correction (FEC) schemes have been deployed in various applications based on Real-Time Transport Protocol (RTP). Since all of them are designed to be used with a certain specific protocol or set of applications, their use is limited, and each application requiring FEC protection must implement the system from scratch....[ Read more ]

This thesis presents the design and implementation of a novel technique in providing quality of service in transferring multimedia data over the Internet. With the advent of IP multicast and increased use of real-time multimedia applications on the Internet which rule out the use of Transmission Control Protocol (TCP), high packet loss rates have been observed in non-TCP protocols. A number of forward error correction (FEC) schemes have been deployed in various applications based on Real-Time Transport Protocol (RTP). Since all of them are designed to be used with a certain specific protocol or set of applications, their use is limited, and each application requiring FEC protection must implement the system from scratch.

In this thesis, a novel approach to protecting packet losses in an Internet Protocol (IP) or User Datagram Protocol (UDP) packet stream called Sub-Socket Forward Error Correction (SSF) has been proposed. Implemented below the socket interface, the system provides transparent FEC protection to any IP-based protocols. This property makes it usable without requiring changes to either the application software or the kernel.

The prototype system uses an erasure code based on the Vandermonde matrix, which can be computed efficiently. Various levels of protection can be applied with varying the coding rates. The packet loss rate can be kept below a desired level by observing the packet loss rate inherent to the channel, and setting the coding rate to the suitable level. The prototype system can also dynamically adjust the coding rate in response to the changing channel condition to satisfy a target packet drop rate. In this way, the loss rate seen by the application can be kept below a given level at all time.

Since the system is light-weight and does not require retransmission, the technique is particularly useful for multimedia applications with real-time constraints. The design, implementation and experimental results of the proposed system are presented in this thesis.