Improving Content and Service Distribution beyond Infrastructure Upgrades

Over the past decades, the Internet has evolved into an essential platform for communication and access to information. This success manifests itself in the steadily growing traffic due to rising user demands for content and services. However, these demands put pressure on operators to improve performance and availability. For instance, Internet Service Providers perform network upgrades to increase bandwidth and Content Distribution Networks operate global infrastructures to move resources closer to users.Despite these efforts, there are limits to infrastructure upgrades and their impact, ranging from technical aspects to economic decisions by the operators.

To still achieve performance improvements in light of the constant Internet evolution, research also focuses on the optimization of aspects other than complete infrastructure upgrades. These efforts include optimizing Internet protocols, better utilization of existing infrastructures, and the development of approaches to overcome limited infrastructure support. However, these optimizations can have unknown consequences in real-world deployments. Exemplary, even if protocols underwent a rigorous standardization process, the Internet-wide adoption, configu- ration, and impact are often unpredictable. The same holds for approaches that try to complement and support current infrastructures, which face practical challenges when deployed outside of testbeds, such as missing acceptance by users due to the lack of control and complex configurations. Overcoming these challenges is relevant to achieve performance improvements beyond the current state-of-the-art solutions.

In this dissertation, we approach the problem of how to provide improvements beyond infrastructure upgrades from the aforesaid two viewpoints: the deployment of a new protocol and the applicability of complementing and supporting approaches.
With regard to protocols, we provide the first large-scale adoption study of HTTP/2, as the new standard protocol for the Web. We analyze how the fundamentally new Server Push feature is utilized, as it promises improvements by reducing the number of messages between a client and a server. Compared to the adoption of HTTP/2, Server Push is only sparsely used and we observe that it does not always lead to improvements, but also to human-perceivable detrimental effects, which we verify in a user study. Building on this, we further inspect which factors influence the performance of Server Push. Based on these insights, we provide a novel strategy for the use of this feature, which can lead to significant performance improvements.
Focusing on approaches that support or complement infrastructures, we present an approach that enables mobile users to connect to private Access Points, using provider-assisted authentication and the isolation in an on-demand created single- purpose network. By that, we provide connectivity in case other infrastructures, such as cellular networks or public wireless hotspots, are currently not available. Finally, we present three systems that realize the distribution of content and services between devices directly, where infrastructure support is only required during setup or tunable reporting and management intervals. We provide means for the manage- ment and control in these systems, as our main goal is their real-world applicability.

In summary, the contributions of this dissertation provide valuable insights and practical technical solutions that complement infrastructure-based improvements.