3.6 Conclusions
Focusing on the relative differentiated version of the IP DiffServ model, we have presented a futuristic QoS mapping framework in this chapter. We have proposed RPI-based video categorization and effective QoS mapping under a given cost constraint. The RPI plays a good bridging role in enabling the network to be content-aware and provides delivery of packets with QoS commensurate with their contents. This results in better end-to-end video quality at a given cost. We have also suggested practical guidelines for effective QoS mapping based on categorized RPI. The performance of content-dependent differentiation was demonstrated by extensive experimental results under the two-state Markov chain, modelling a different random packet loss rate for each network DS level (DiffServ class). The results clearly indicate the benefit of the proposed mapping framework. Experimental results showed that the proposed framework enhances end-to-end video quality at the same total cost.
Two areas in this proposed framework can be further elaborated. First, only the differentiation of loss rates was performed in this work, and it must be extended to cover another key QoS measure—delay/jitter. Loss rate/delay-combined differentiation will provide a more comprehensive characterization of multimedia contents (for the video stream itself as well as among various kinds of multimedia traffic). Thus, by summarizing loss rate and delay priorities in the DS byte of the packet header, more enhanced, content-dependent forwarding will be feasible. Second, to consider a more accurate evaluation of error propagation for the QoS mapping decision, we assigned RLIs to source packets on the basis of the corruption model in Section 3.3.2. Such schemes further improve performance. However, assignment based on the simple online-generated RLI in Section 3.3.3 has the advantage of online implementability and simplicity. Indeed, QoS mapping based on this simple online RLI assignment works fairly well already.