Internet traffic reports show that YouTube is one of the major sources of data traffic world-wide. Furthermore, the data traffic shifts from mostly fixed landlines to cellular data connections where bandwidth is sparse and expensive. Previous studies revealed that YouTube uses a user-friendly HTTP Adaptive Streaming (HAS) strategy which sacrifices bandwidth efficiency to increase the average playback quality for the user. That way, it happens that the same video segment is transmitted in two or more quality levels, but only one can be shown to the user. We denote this as redundant traffic and this work is dedicated to understanding the influence factors on the amount of redundant traffic. This paper presents the results of a large-scale study with over 12,000 video views over a bottleneck link shaped to various bandwidths.We first evaluate the playback characteristics and show that YouTube’s HAS algorithm linearly increases the average playback quality with the available bandwidth while at the same time video buffering is sublinearly decreased. Furthermore, we identify video-dependent bandwidths which optimize the playback time on a quality level. Afterward, we show that this is achieved by discarding lower layer segments and therefore paid with redundant traffic of up to 40%. We evaluate the overall efficiency of the system and show that YouTube is able to improve the average quality level by up to 0.7 quality levels by using this adaptation strategy. However, a penalty of 0.5 quality levels is paid for it due to the discarded data of the lower quality segments.

YouTube is one of the most popular and volume-dominant services in today’s Internet, and has changed the Web for ever. Consequently, network operators are forced to consider it in the design, deployment, and optimization of their networks. Taming YouTube requires a good understanding of the complete YouTube stack, from the network streaming service to the application itself. Understanding the interplays between individual YouTube functionalities and their implications for traffic and user Quality of Experience (QoE) becomes paramount nowadays. In this paper we characterize and model the YouTube stack at different layers, going from the generated network traffic to the QoE perceived by the users watching YouTube videos. Firstly, we present a network traffic model for the YouTube flow control mechanism, which permits to understand how YouTube provisions video traffic flows to users. Secondly, we investigate how traffic is consumed at the client side, deriving a simple model for the YouTube application. Thirdly, we analyze the implications for the end user, and present a model for the quality as perceived by them. This model is finally integrated into a system for real time QoE-based YouTube monitoring, highly useful to operators to assess the performance of their networks for provisioning YouTube videos. The central parameter for all the presented models is the buffer level at the YouTube application layer. This paper provides an extensive compendium of objective tools and models for network operators to better understand the YouTube traffic in their networks, to predict the playback behavior of the video player, and to assess how well they are doing in practice in delivering YouTube videos to their customers.

This report documents the program and the outcomes of Dagstuhl Seminar 15481 "Evaluation in the Crowd: Crowdsourcing and Human-Centred Experiments". Human-centred empirical evaluations play important roles in the fields of human-computer interaction, visualization, graphics, multimedia, and psychology. The advent of crowdsourcing platforms, such as Amazon Mechanical Turk or Microworkers, has provided a revolutionary methodology to conduct human-centred experiments. Through such platforms, experiments can now collect data from hundreds, even thousands, of participants from a diverse user community over a matter of weeks, greatly increasing the ease with which we can collect data as well as the power and generalizability of experimental results. However, such an experimental platform does not come without its problems: ensuring participant investment in the task, defining experimental controls, and understanding the ethics behind deploying such experiments en-masse. The major interests of the seminar participants were focused in different working groups on (W1) Crowdsourcing Technology, (W2) Crowdsourcing Community, (W3) Crowdsourcing vs. Lab, (W4) Crowdsourcing & Visualization, (W5) Crowdsourcing & Psychology, (W6) Crowdsourcing & QoE Assessment.

To cope with the increasing demand of mobile devices and the limited capacity of cellular networks mobile connections are offloaded to WiFi. The access capacity is further increased, by aggregating bandwidth of WiFi access links. To analyze the performance of aggregated access links we model the most simple case of two cooperating systems interchanging capacities using an offloading scheme. The resulting analytic model is computed by means of a two-dimensional birth and death process. It can be used to seamlessly evaluate the performance of systems between partitioning and complete sharing. This allows to optimize the setting of thresholds dependent on the load of the cooperating system. Furthermore the benefit of aggregating bandwidth in different scenarios with homogeneous and heterogeneous workloads is quantified and the performance of more than two cooperating systems is evaluated by simulation.

HTTP Adaptive Streaming (HAS) technologies, e.g., Apple HLS or MPEG-DASH, automatically adapt the delivered video quality to the available network. This reduces stalling of the video but additionally introduces quality switches, which also influence the user-perceived Quality of Experience (QoE). In this work, we conduct a subjective study to identify the impact of adaptation parameters on QoE. The results indicate that the video quality has to be maximized first, and that the number of quality switches is less important. Based on these results, a method to compute the optimal QoE-optimal adaptation strategy for HAS on a per user basis with mixed-integer linear programming is presented. This QoE-optimal adaptation enables the benchmarking of existing adaptation algorithms for any given network condition. Moreover, the investigated concept is extended to a multi-user IPTV scenario. The question is answered whether video quality, and thereby, the QoE can be shared in a fair manner among the involved users.

This report documents the program and the outcomes of Dagstuhl Seminar 13361 "Crowdsourcing: From Theory to Practice and Long-Term Perspectives". Crowdsourcing is a newly emerging service platform and business model in the Internet. In contrast to outsourcing, where a job is performed by a designated worker or employee, crowdsourcing means to outsource a job to a large, anonymous crowd of workers, the so-called human cloud, in the form of an open call. Current research in crowdsourcing addresses the following issues: crowdsourcing as a novel methodology for user-centered research; development of new services and applications based on human sensing, computation, and problem solving; engineering of improved crowdsourcing platforms including quality control mechanisms; incentive design and gamification of work; usage of crowdsourcing for professional business; theoretical frameworks for evaluation. The topic on crowdsourcing may have a huge impact on the Internet and its technical infrastructure, on society, and the future of work. In short, crowdsourcing will be a guiding paradigm and form the evolution of work in the next years. Therefore, this seminar helps coordinating research efforts in the different communities. In five presentation and discussion sessions, the diverse aspects of crowdsourcing were elaborated. The topics of the sessions covered (S1) crowdsourcing in general, (S2) industry use cases, (S3) crowdsourcing design and engineering, (S4) programming and implementing crowdsourcing, (S5) applications of crowdsourcing. The major interests of the seminar participants were then focused in four different working groups on (W1) long-term perspectives & impact on economics in five years, (W2) theory -- taxonomy and dimensions of crowdsourcing, (W3) industry use cases, (W4) crowdsourcing mechanisms and design. In parallel to this seminar, a topically related seminar on "Cloud-based Software Crowdsouring", organized by Michael N. Huhns, Wei Li, Martin Schader and Wei-Tek Tsal,(Dagstuhl Seminar 13362) took place. Therefore, a joint late night session was organized to discuss crowdsourcing with respect to ethics and its relation to social computation.