Most resource allocation decisions in converged multi-service networks are presently driven by QoS parameters and service level agreements. Network operators are realizing that such policy decisions ought to be based on Quality of Experience (QoE), which is a better measure of subjective video perception. Video QoE has two components: (a) Zapping and (b) Video Quality or Video Mean Opinion Score (MOS). Zapping anomalies are well understood and easily characterized, but there is no consensus on what the definition of a Video-MOS should be. This is largely because there is little understanding of qualitative and quantitative subjective degradations caused by networking events like loss, delay, jitter, and error. In this work we conduct extensive experiments on a simulation testbed in a search for a good Video-MOS definition, with specific emphasis on quality degradations due to network transmission. We deploy "collector" nodes that can gather various statistics of a given flow. We also identify key parameters that such collector nodes must maintain such that an overlay of such nodes can identity impairment points. Lightweight Video-MOS definition evolving out this work can be used in collector overlay networks to solve problems like root-cause analysis, capacity-planning and various network optimization problems.
展开▼
