Abstract
The demand for over-the-top (OTT) video delivery (also known as video delivery over the Internet) is increasing rapidly. According to the statistics gathered by Cisco, IP video traffic will be 82% of all global IP traffic by 2022, up from 75% in 2017.
Our contributions within this thesis focus on two OTT video delivery methods—HTTP-based Adaptive Streaming (HAS) and HTTP progressive download. Specifically, we focus on the issues of the aforementioned video delivery methods that can degrade the end-users' QoE, and propose solutions to mitigate those problems with the goal of improving the perceived experience of the end-users.
In HAS, at the server-side, a given video is available in multiple quality levels (in different bitrates and resolutions). The client-side player (HAS player) can measure the available bandwidth with the server, and dynamically adjust the requested quality to match with the underlying network conditions. This feature will significantly help the HAS players to avoid re-buffering events (playback interruptions) that can occur due to the fluctuating throughput provided by the best-effort Internet. Within the space of HAS, we examine the HAS multi-client competition problem—when two or more HAS players compete for network bandwidth through a bottleneck link, e.g., on the access link of a home network or an enterprise LAN. This competition can lead to two critical problems in terms of the end-users' perception of QoE—instability in the quality level of the playback, and unfairness between different players' quality levels. Primarily, this happens because the competing players often fail to determine the correct fair share of the bandwidth available to them. To this end, we propose Coordinated-DASH (CDASH), a lightweight coordination scheme that strives to enhance the end-users' QoE by reducing the instability and unfairness issues as mentioned above. Our experiments show CDASH to reduce instability at least by 25% but often up to 71%, and unfairness by no less than 35% to a maximum of 83%, compared to state-of-the-art.
Unlike HAS players, players that support HTTP progressive download cannot dynamically change the requested quality to adapt to the underlying network changes. This limitation makes the progressive download players more vulnerable to the re-buffering events that can negatively affect the QoE of the end-users. One straightforward solution to this issue is to maintain a larger playback buffer at the client-side player. A larger buffer generally provides stronger protection against re-buffering events, since it accommodates a broader range of download rate variability. However, larger buffers also incur longer start-up delays, and longer delays in refilling, should a re-buffering event occur. On the other hand, longer start-up and re-buffering delays are also shown to be degrading the end-users' QoE. Aiming to handle this trade-off effectively, we propose an adaptive playback buffering scheme for progressive download players—Variable bitrate Pattern Aware Playback Buffering (VPAP). VPAP dynamically calculates the minimum required buffer size per streaming session that will allow the player to start playback as early as possible while reducing the likelihood of re-buffering events during playback. According to the experimental results, in the best case, VPAP was able to start playback with less than 1% of the delay incurred by state-of-the-art solutions, while still avoiding re-buffering events. In the worst case, when state-of-the-art solutions had more than six re-buffering events, VPAP reduced it to zero with the expense of an increased start-up delay.