The early stages of video delivery relied on downloading, i.e., the user saves the full video on its computer and then watches it. Content providers (CP) then realized that the user quality of experience (QoE) would be improved if the user could watch parts of the video as soon as it is available in the user equipment (UE), i.e., the video is downloaded and played at the same time (streaming).
To guarantee smooth streaming, the UE must possess a connection that provides a bandwidth equal to or higher than the video bitrate. Otherwise, the player will not acquire content fast enough and cause a stalling event, which reduces the QoE.
In general, users prefer a lower video quality than a stalling event, to this end, CPs encode the videos in different quality levels, and the better the quality, the higher the required bitrate. With these different video qualities available, the user application can select which quality his connection can support without stalling. This strategy enables CP to provide a good service despite the diversity of connections available to different users.
These strategies do not solve the whole problem. One difficulty which is not tackled by the presented approach is that in real networks the UE available bandwidth also changes over time, so an estimated bitrate cannot be assumed to the entire duration of the video. This bandwidth instability is even more present in mobile networks.
The solution to cope with the variation of bandwidth is to vary also the video bitrate. Hence, the state of the art solution divides the video into small segments (seconds) so that the user application can choose which quality to present to the user on a segment basis. This strategy allows the UE to dynamically adapt to variations in the available bandwidth and avoid stalling, at the cost of decreased video quality, and thus, provide a better QoE.
This delivery, as seen from the operator side, is done over an encrypted channel, consequently posing at least two significant challenges to network operators that want to optimize the network focused on delivering the best QoE to users. Firstly, the network operator does not even know if the user is requesting a video and thus has no information of the application layer. Secondly, even if the network operator knows that the user is streaming a video, the streaming quality is still hidden. One example is that with adaptive streaming network congestion can be hidden from the network operator, given that on congestion, the client adapts to a lower quality, thus alleviating the network and leading the operator to think that the client is experiencing a reasonable service, but it is actually playing a bad quality video.
The SPOTLIGHT project is investigating new technologies to improve the described problem, i.e., to allow network operators to optimize the delivery of encrypted adaptive video streaming.