Scaleable and Smooth TCP-Friendly Receiver-Based Layered Multicast Protocol

In the presence of heterogeneity and scalability, i.e. the case when delivering real-time television content over the Internet, receiver-based layered multicast communication is the most efficient way to deliver real-time video data to a large number of receivers. TCP-friendly Receiver-based Layered Multicast Protocol (TRLMP) is a protocol that supports layered multicast real-time video delivery, while maintaining the fair sharing of the network resources. However, there is room for improvement on the current TRLMPs. Some of the current TRLMPs and their techniques are not efficient in supporting the delivery of real-time video data on the Internet, in particular when the number of receivers is very large. The current TRLMPs' techniques do not allow the receivers to: 1) estimate Round Trip Time (RTT) in a scalable manner; and 2) achieve a steady reception rate because of volatile loss rate estimation. These problems result in stability and scalability problems for TRLMPs. In this thesis, we propose a new TRLMP called Scalable, Smooth and TCP-friendly Receiver-Based Layered Multicast Protocol (SS-TRLMP) to address the problems of the current TRLMP. Similar to other TRLMPs, SS-TRLMP relies on a TCP-equation model to control congestion and fair behaviour of the protocol. However, the equation model requires RTT data, which is problematic to estimate when the number of receivers is too large. In order to address this problem, we proposed the Scalable RTT Estimation Technique (SRTT) for layered multicast communication. The technique models Transmission Control Protocol (TCP) behaviour more accurately than the current RTT estimation techniques. Moreover, the technique is simple and can be easily implemented in the current Internet infrastructure. Stability is one of the requirements of real-time video applications. However, even with the current loss rate smoothing techniques, TRLMPs behaviour is highly volatile. Moreover, in layered multicast communication there is no synchronisation of packets between multicast layers, which causes misinterpretation of loss events at receivers. We address these problems with packet re-ordering at the receiver and the 2-Step Loss Filtering Technique. Packet reordering technique enables the TRLMP to estimate accurate loss rate similar to TCP, and 2STEPS provides better stability and responsiveness than the current techniques. Finally, by combining the techniques of the current TRLMPs and the proposed techniques, we designed SS-TRLMP. The proposed SS-TRLMP has the following properties: TCP fairness, scalability and stability