Wenzel, Georg; Gehrke, Gerhard; Stoll, Dieter; Thomas, Wolfgang (Alcatel-Lucent Deutschland AG, Thurn & Taxisstr. 10, 90411 Nuremberg, Germany)
Belzner, Martin (Chair of Information Transmission, Friedrich-Alexander University Erlangen-Nuremberg, Cauerstr. 7, D-91058 Erlangen, Germany)
It is already known that significance marking of IP video packets in conjunction with a corresponding evaluation in the transport network considerably protects the quality of videos against degradation from network congestion. The focus of this paper therefore is the investigation of possibilities to take benefit from such significance marks. In particular the interest is to estimate the video quality in the network without having to spend considerable effort for decoding or even decrypting of the video stream. While the merit of significance markings has already been proven by simulation, now the benefits have been demonstrated by experiments on actual transport network hardware. Thereby, network element typical Ethernet performance counters registering the amount of received and lost bytes have been modified for differentiation by the significance marks. Those counters, retrieved per second, were compared with the quality of the video. It is discussed how to construct a metric for video quality, and the role of a suitable definition of an objective reference for the calibration of this metric: The metric has to regard specifics of the video compression, video encoding and transport network characteristics. Results from these experiments are provided and challenges for further refinement and implementation are pointed out.