Defence of doctoral thesis in the field of networking technology, M.Sc. Alemnew Asrese
The web has been an integral part of our daily lives. For the last three decades, it has been evolving dramatically and has become a complex ecosystem on the Internet. This evolution in the web ecosystem has not been limited to the contents but also includes various advancements aiming at improving the web performance and the user experience. Measuring users' browsing experience requires a complex design of user studies to assess the perceived quality of the service under study. Conducting a user study is time and resource consuming and hence not a scalable solution. Thus, it is important to design metrics and methodology along with automated tools to assess the quality of experience of services from systems themselves, e.g., using active network measurement.
This thesis presents metrics, methods, and tools for measuring web quality of experience (QoE) at scale. Furthermore, the thesis presents the analysis of datasets collected over a long time period using the tools applied to both fixed-line and cellular networks. In the thesis, machine learning models are used to map the objective metrics to the subjective ratings obtained in user studies. The thesis contributes research to make automated models to predict the human perception of the web page loading process. The metrics, methods, tools, and datasets presented in this thesis are relevant for Internet practitioners and web engineers.
The results in the thesis show time-based web performance metrics are impacted by several factors including the complexity of web pages, the availability of caches to download the content from a closer location. For popular web pages such as Google and YouTube, web page loading times have improved over time. The measurement also confirmed that an improvement in the broadband speed does not necessarily yield faster loading time for web pages and better browsing experiences. The results further showed that roaming shorter distances (e.g. from Finland to Sweden) does not have an impact on the user's browsing experience.
The software and the results of this thesis will help web designers, content, and service providers to better design and manage infrastructures to improve the end-user experience. ISPs can deploy small measurement boxes at the customers’ home gateway, and execute measurement tools to monitor the web QoE from the customers’ location. The ISP can also use our measurement results for better capacity planning and network design.
Opponent: associate professor Gareth Tyson, Queen Mary University of London, Great Britain
Custos: professor Jörg Ott, Aalto University School of Electrical Engineering, Department of Communications and Networking
Contact information of doctoral candidate: Alemnew Shefera Asrese, [email protected], +358456987878
The defence will ne organised via remote technology (Zoom). Link to the defence
The doctoral thesis will be publicly displayed 10 days before the defence in the Aaltodoc publication archive of Aalto University.