Bandwidth allocation optimization for IPTV QoS/QoE prediction & measurement

Abstract

There are two different aspects, i.e. ``network`` and ``multimedia,`` that are both closely coupled in many critical issues such as QoS, QoE, etc., for IPTV and mobile IPTV services. An important problem is to provide realtime QoS/QoE-guaranteed multimedia services over packet-based networks. The problem is still unsolved because there are many parameters affecting quality between network and multimedia. In order to solve the problem, a study on QoS/QoE prediction and measurement is necessary. IP networks are on a steep slope of innovation that will make them the longterm carrier of all types of traffic, including multimedia for IPTV and mobile IPTV services in the NGN environment today. However, such networks are not designed to support realtime multimedia communication because their variable characteristics (e.g. due to bandwidth, packet loss, delay, etc.) lead to a deterioration in video quality. A major challenge in such networks is how to predict or measure video quality accurately and efficiently considering network resources that provide QoS/QoE-guaranteed services. Finally, we propose QoS/QoE prediction and assessment methodology by model design and performance analysis in network and in video. For the QoS prediction in the network level, we propose and analyze resource allocation algorithms based on reward optimization with bandwidth allocation of four different class levels in a constraints limited system resource. First of all, the formulated bandwidth allocation and adaptation algorithms for real-time QoS-guaranteed multimedia service in this research are useful to get an advantage in cost benefit as an aspect of the system or service provider. For QoS requesting and delivering in realtime, these bandwidth allocation and adaptation algorithms are developed based on two strategies: cost-based admission control in priority reservation regarding system (service provider) benefit and flow-based dynamic adaptation control regarding user satisfy in Q...