Video Streaming Quality of Experience (QoE): In-network Cache Prefetching and Moving QoE Models
No Thumbnail Available
Authors
Meeting name
Sponsors
Date
Journal Title
Format
Subject
Abstract
Video streaming accounts for a significant amount of traffic on the Internet. Users expect a high quality of experience with online video streaming. Service and content providers desire to provide a satisfying experience for end users. Therefore, developing metrics to measure users satisfaction of such services is crucial. Quality of Experience (QoE) of a user for a video streaming service is important for content providers. For video streaming, the DASH (Dynamic Adaptive Streaming over HTTP) standard is one of the common approaches for streaming used by content providers in which a video is divided into segments of different bit rates for delivery. In this research, we studied two inter-related problems for DASH video streaming: 1) in-network caching techniques, prefetching, and 2) QoE measurement and monitoring. In the first research direction, we focus on content providers utilizing in-network caching and prefetching in order to reduce video delivery latency to provide users a higher quality of experience and reduce the traffic load on the core network. The issue with current prefetching methods is that they do not utilize available resources well; thus, the end users are not able to receive the best possible QoE. These approaches are either mostly naive or they are not compatible with the DASH protocol and they are too complex consuming too much time and compute resources. We propose a smart video cache prefetching scheme for segment bitrates. Our prefetching approach is based on throughput values in the cache that are forecasted using previous throughput values from clients. Since in a cache environment, multiple clients contend for video segments in the cache, we assess the cache performance and also con- sider the impact on QoE for each client during contention. When comparing our scheme with an existing scheme, results show that our smart prefetching increases the cache hitrate and reduces the number of unused prefetches for the cache, thereby improving QoE of the clients. In our second research direction, we focus on objective QoE, for which a number of QoE models has been proposed. The limitations of the current models are that the QoE is provided after the entire video is delivered; also, the models are on a per client basis. We refer to such models as static QoE models. In many situations, such as live events, ensemble QoE during the session is important to understand, especially for multiple clients together, for network and content providers. For this need, we propose two QoE models to capture QoE periodically during video streaming by multiple clients simultaneously, which we refer to as Moving QoE (MQoE) models. Our first model, MQoE_RF takes into consideration the nonlinear effect due to the bitrate gain and sensitivity from the bitrate switching frequency. Our second model, MQoE_SD focuses on capturing the standard deviation in the bitrate switching magnitude among segments. Then, we study the effectiveness of both the models in a multi-user mobile client environment. We compared our models with an extension of the Model Predictive Control (MPC) QoE model (referred to as MQoE_MO). Our study shows the robustness of our MQoE models. The results show how the MQoE models is able to more accurately capture the overall QoE behavior than the static QoE model and its extension.
Table of Contents
Introduction -- Literature survey -- Smart Cache Prefetching -- Smart Cache Prefetching evaluation -- Moving QoE Models -- Moving QoE Models evaluation -- Conclusion
DOI
PubMed ID
Degree
Ph.D. (Doctor of Philosophy)