What are the most used metrics to evaluate the QoS of a video streaming session? I am looking for tutorials, papers, lecture notes, ppt, anything of that kind, or your own answer.
From the perceptual front (in the user point of view):
Quality: Here we are evaluating quality of video and audio being streamed. Usually, higher the bit rate better the quality - but it also means more cost. The basic quality measure for Video is PSNR. However, more relevant matrices from the perceptual quality is blockiness and there are many other approaches like this. Usually, choosing the codec and bit rate fixes the quality primarily, so network doesn't effect that much in that case.
Consistency (or lack of jerkiness): This is a rough measure of times when the continually running stream got interrupted. Usually this is a measure of length and pauses that are caused due to unavailability or reduction of bandwidth during some time and player needs to wait for the data to arrive. Usually, buffering is done to the optimal level in the receiver to minimize the effect on this.
Errors in pictures: This is the indication of errors occurred during the transmission due to which patchy or blank screens. There are cases, where transmission could be error free, in which case, this won't be a problem.
Latency: This refer to amount of time the service needs to begin the deliver and time the system gets reflected to screen. There are various elements of the system that contributes to this. Usually, for VoD type systems, some reasonable but it is a reasonable concern for the two systems.
End-to-end delay or so called guarantee: The end-to-end delay is essentially same as round trip delay between two systems. For two way communications like telephony, end-to-end delay is of great importance.
Service guarantee: Usually, this refers to probability with which the given service will be surely available to the end users. This is equivalent to probability of call drops or network blocking of calls.
Here we can think of QoS metrics from the point of view of basic measures of the network which is reflects the QoS as perceived by user.
Bit rate guarantees: This includes the maximum and minimum of available bandwidth during the entire session. Guarantees available for the network bandwidth is usually the single most
Error rate or Loss runs: Usually, unlike satellite paths, in the internet there are no bit errors (above transport layer) - the errors get reflected in the form of packet losses. Basic metrics is hence packet error rate on the networks. More research has concluded that these can be modeled as loss lengths where typically, errors occur in some bursts. The packet loss and the length of the losses has a significant effect in terms of perceptions. This paper describes this metric in depth.
Latencies or Various delays: This involves the delay between the path, setup delay, resource reservation delays and buffer delay that all contributes to initialization of display and latency. Many a times, efficient codecs collects various pictures together before encoding to optimize bit rates this allows more efficient compression (higher quality) but increases the latency.
System uptime and usage factor: If either the systems is dysfunctional (down) or it is fully utilized (blocked) it will have
The paper below shows good relationship between various taxonomies. Survey of QoS specifications : http://eia.udg.es/~marzo/doctorat/aur98.pdf