Service Quality Improvement and User Experience Optimization by Introducing Intelligence in the Network

Abstract

The majority of present-day distributed applications demand a certain level of service from the transportation network and impose a number of performance requirements on it. Failure to meet these requirements will typically degrade the efficiency of the application and, more importantly, will likely have a detrimental impact on the experience of the end-user. Over the years, a number of frameworks has therefore been proposed that enable transportation networks to guarantee a certain Quality of Service (QoS) and the positive implications of these frameworks on application performance have been established. Due to the growing attention to human factors however, the focus is increasingly shifting from pure QoS provision towards user experience optimization. Stated differently, in recent years, the goal of guaranteeing a certain level of network performance is evermore being replaced with the aspiration to ensure a high end-user satisfaction or so-called Quality of Experience (QoE). The current generation of telecommunications networks, and the omnipresent Internet in particular, unfortunately lack elaborate and effective constructs for QoE optimization. Mitigating this deficiency forms the subject of this thesis.

The overall contribution of this doctoral dissertation is the Network Intelligence Proxy (NIProxy), a network intermediary which has been developed to enable QoE manipulation and optimization in IPv4-based computer networks. As is hinted at by its name, the NIProxy's methodology involves the introduction of "intelligence" in the networking infrastructure. This is achieved by accumulating contextual knowledge regarding the transportation network itself, the distributed application and the end-user (and his terminal).

On a finer-grained level, this PhD research contributes to various domains, the first of which being network traffic engineering and application-layer QoS provision. The NIProxy provides two complementary techniques to improve the (multimedia) traffic handling capabilities of the transportation network in which it is deployed. The first technique, network traffic shaping (NTS), enables the in-network coordination and management of the bandwidth consumption of the network traffic that is induced by distributed applications. The second tool, service provision, allows the NIProxy to act as a substrate for the hosting of services that will be applied to the (multimedia) content that fl ows through the network. Both traffic engineering mechanisms are context-aware and context-adaptive since they have unbridled access to the NIProxy's context repository.

A second concrete research contribution is situated in the area of network intermediary design and implementation. The service provision platform is designed with extensibility in mind. Thanks to a pluggable implementation, the set of provided services can at run-time be modified and extended. This extensible design for instance enables the NIProxy to cope with heterogeneous conditions in dynamic networking environments without requiring a reboot. Secondly, instead of implementing the NTS and service provision frameworks as isolated entities, an integrated design is adopted so that both traffic engineering techniques are allowed to interface and collaborate. The outcome is a holistic solution in which the composing components supplement each other's strong points and improve their potential through cooperation.

Finally, this PhD dissertation presents significant results in the area of context-aware networking and user experience optimization. The ultimate objective of the NIProxy is not to improve the QoS features of the transportation network, but instead to improve the experience of the users which leverage the network to run distributed applications. The viability and the feasibility of the NIProxy as a QoE manipulation and optimization framework is investigated by analyzing the outcomes of numerous experimental evaluations. These studies confirm that the NIProxy is equipped with an elaborate toolset for user experience optimization which enables it to deliver on its objective in a multitude of distributed scenarios, in dynamic environments, under variable network load and amidst heterogeneous mixes of network traffic types. It is also corroborated experimentally that the NIProxy is amenable to cooperation with other QoS provision or QoE optimization solutions and as such can serve as a building block for the construction of larger frameworks.

To conclude, it is worth underscoring what this dissertation is not about. The NIProxy is not concerned with (methods for) QoE measurement or the QoE concept per se. This thesis also does not involve the collection of qualitative feedback by means of user studies. Lastly, the NIProxy is not a ready-made solution for QoE optimization; instead, it is a QoE manipulation framework whose actual behavior needs to be attuned to the current context of use. Simply integrating the NIProxy in a transportation network is hence by no means a guarantee for success.