Affiliations: Mathematical and Computer Science Department,
Jacksonville State University, AL, USA | College of Engineering and Computer Science, Wright
State University, OH, USA
Abstract: In the best-effort IP network, quality of service (QoS) cannot be
guaranteed, and thus packets could possibly be delayed or lost. Packet
delay/loss will inevitably degrade the perceptual quality of real-time
multimedia-over-IP service, such as Voice-over-IP (VoIP), Video-on-Demand
(VoD), etc. In general, packet loss/delay exhibits temporal dependence. In
order to efficiently conduct error recovery/concealment and improve the
perceptual quality of the transmitted multimedia contents, packet loss/delay
has to be precisely modeled. Different mathematical models, such as Bernoulli
Model, Gilbert Model, Extended Gilbert Model, have been proposed to model
network trace. However, none of them is able to precisely model the networked
multimedia trace like the General Markov Model (GMM), which has rarely been
applied in practice due to massive computational resource requirements. With
the developments of modern computing hardware and parallel processing
algorithms, GMM is becoming computationally feasible. In this paper, in order
to obtain network traces for real-time VoD transmission, different connections
have been setup to simulate a real VoD system. Data packets have been
transmitted between the server and clients under RTP/UDP/IP protocol stack.
Different models have been applied to analyze and model the obtained video
transmission network traces. Specifically, a 6-state GMM has been applied to
analyze the network trace, and the parameterized model has been obtained for
further error recover/concealment. Compared to the other models, GMM offers the
best modeling precision, in terms of loss-run distribution (LSD) and Forward
Error Correction (FEC) performance prediction. The parameterized GMM is very
useful to model and analyze network traces and further improve the QoS in
multimedia-over-IP based on the modeling and analysis.
Keywords: modelling, simulation, network traces, video transmission, IP networks
