User experience and network resource optimization for live video streaming
Recent social media trends have proliferated the demand for HD live streaming events over the Internet. Such events may include sports, TV channels or individuals broadcasting to an audience located across the globe. In live video streaming, multiple users subscribe to the same event at the same time, increasing the peak bandwidth requirements. Consequently, a dynamic and flexible provisioning of network and system resources becomes difficult. With user preferences shifting towards mobile devices, cellular network operators also face challenges when handling flash-crowds viewing live video. The rising adaptation of SDN by ISP and CDN presents an opportunity to dynamically respond to short-lived broadcast events as well as HD mega events. SDN's centralized control is utilized to propose mCast, an architecture that enables inter-domain network-layer multicast and avoids redundant transmissions in core and wired access networks. To handle network congestion, Danos is proposed, that deploys an optimization model to maximize the user video quality while minimizing the resource consumption. The model considers device capabilities, network constraints and user’s ISP or CDN subscription levels. In the cellular domain, the standard eMBMS improves the utilization of scarce wireless resources. Two key decisions when configuring an eMBMS service area are: which eNB to synchronize and how to share resources among users with heterogeneous channel conditions. To optimize network configuration, first RTOP is proposed. RTOP formulates a joint optimization model and employs a real-time heuristics-based algorithm for a single network instance. Then NIMBLE is proposed to address design challenges that arise due to temporal aspects of network and user-state. NIMBLE reacts to variability and re-configures the network to increase user QoE. Large-scale test-beds were developed to compare the proposed algorithms with state of the art approaches and various network and video performance metrics were evaluated. Results showed a 70% increase in average user throughput and elimination of frame drops in core and access network. Similarly, the eMBMS algorithms increased average throughput by 150% and reduced the bitrate switches by 75%. Overall, 80% of the users had an improved QoE.
eMBMS , Quality of experience , Software defined networking , Optimization models , Multimedia streaming
Khalid, A. 2019. User experience and network resource optimization for live video streaming. PhD Thesis, University College Cork.