YouTube QoE-Aware Gateway Selection in Future Wireless Networks

The evolution and growing dissemination of the wireless Internet as well as the considerable usage heterogeneity introduced by new fields of applications, raise the need for intelligent resource management in future wireless networks. This necessity is urgent as state-of-the-art wireless resource management is mostly (1) unaware of the Internet applications in the network, (2) enables best effort Internet access only and (3) does not allow application-network interaction. This has been identified as a serious drawback for modern wireless networks and results in different research activities aiming at developing cross-layer and network wide control and management concepts [1]. One approach towards a more intelligent resource management is the idea of application comfort (AC) based resource management [2], [3]. Application comfort quantifies how well an application is currently running and allows a prediction of the quality of experience (QoE). An AC monitoring tool running at the client and cooperating with resource management tools thus allows (1) to guarantee application-specific service levels and (2) to react prior to a QoE degradation. In contrast to conventional solutions, the use of an AC monitoring tool avoids the need of complex flow classification techniques like deep packet inspection, TCP/IP or application header analysis or the use of heuristics, as the tool simply signals the existence of a certain application to the network. Secondly, a QoE-based resource management does not need to know applicationspecific QoS requirements which are often not known or even changing over time as it is e.g. the case for videos with variable bite rate, but is able to adapt the network resources in dependence on the application performance. To illustrate the benefits of AC for a network resource management architecture we propose the following demonstration: A client in a wireless mesh network (WMN) is playing YouTube videos while the YouTube AC monitoring tool YoMo [2] continuously monitors the YouTube AC at the client. We chose the example of WMNs since they are an attractive low-cost option for offering broadband wireless Internet access. Second, practically implementing resource