Bushra Anjum

Bandwidth estimation for video streaming under percentile delay, jitter, and packet loss rate constraints using traces

Abstract We present and use a CPU-efficient activity-based simulation model to calculate the sojourn time of a packet and the packet loss rate in a tandem queueing network that depicts the path of a video flow. The video flow is characterized by a packet trace. Background traffic, also characterized by a trace, is allowed in the tandem queueing network. In our analysis we used real video traces (Telepresence, WebEx, Jabber) and also generalized our results using traces generated by a theoretical model of a video arrival process depicted by a Markovian Arrival Process. Using this simulation model we calculate the bandwidth required for a video flow, so that a given set of constraints for the percentile end-to-end delay, jitter, and packet loss rate are satisfied. We also show that the bandwidth required for n identical video streams that follow the same path through an IP network, so that the end-to-end percentile delay remains the same, is a linear function of n . Further, it is experimentally depicted that for infinite-capacity queues the bandwidth required to satisfy the percentile end-to-end delay constraint also satisfies the jitter constraint. And for finite-capacity queues, the bandwidth required to satisfy both the percentile end-to-end delay and the packet loss rate constraints also satisfies the pair of jitter and packet loss rate constraints.

A QoS evaluation of video traffic models for H.264 AVC video

In this paper, we evaluated and compared the QoS behavior of video traffic models for H.264 AVC video. The H.264 AVC models that we evaluated are: the Markov Modulated Gamma (MMG) model, the Discrete Autoregressive (DAR) model, the second order Autoregressive AR(2) model, and a wavelet-based model. These models were used to generate synthetic packet traces which were used in a simulation model to estimate the 95th percentile of the end-to-end delay, jitter and packet loss. The QoS metrics of the generated traces are compared with those of the original traces available from the Arizona State University video traces library. We observed that none of the model produces accurate results for every type of video but the MMG model produced the results closest to the actual traces.

Optimizing channel access for bandwidth versus latency in 802.11ac WLANs

It has been a major criticism to 802.11ac that there are no well-defined channel selection rules when the devices are contending for the secondary 20MHz channel. This leads to inefficient selections. We define efficiency in two distinct ways, latency optimization and bandwidth optimization. We propose a simple greedy strategy of channel assignment for latency optimization, i.e., start assigning stations (STAs) to the highest frequency band they can support and move down from there if that channel is unavailable. For bandwidth optimization, we propose two heuristic based rules for the Medium Access Control (MAC) protocol. A capability based contention window and a bottom up approach towards channel assignment, i.e., start channel contention for the lowest frequency band that a STA can support and jump up to a higher band if contention is high. Detailed simulation studies show that our strategies indeed lead to better latency and bandwidth optimizations.

End-to-end delay percentiles for video traces using a MAP2 approximation

Video traffic is widely expected to account for a large portion of the traffic in future wired and wireless networks. We propose an efficient and accurate approximation method for calculating a given percentile of the end-to-end delay of a video stream depicted by a trace. The queueing delay encountered in the network by the IP packets carrying the video is modeled by a tandem queueing network of infinite capacity queues. The video trace is approximated by a two-stage Markovian Arrival Process (MAP2), which is the arrival process to the tandem network. The proposed method uses only the first queue of the tandem queueing network to construct an upper and lower bound of a given percentile of the end-to-end delay. The percentile value is then approximated by interpolating between the two bounds. We used this method to estimate the 95th percentile of the end-to-end delay of two different types of video traces, Ciscos point-to-point presence and WebEx over a 10-node path. The results obtained were compared against simulation, and have an average relative error of 4.24%. Using this method, we then obtained the minimum amount of bandwidth required to be allocated on each link along the path of the video stream, so that a given 95th percentile of the end-to-end delay is satisfied.

MMPP-2 approximation of aggregate IPPs with application to expedited forwarding class in ipbased networks

In this paper, we investigate the suitability of a two state Markov Modulated Poisson Process (MMPP-2) approximation to the superposition of n homogeneous Interrupted Poisson Process (IPP) streams which pass through a series of m homogeneous network nodes. Earlier, such approximations have been attempted with the following restrictions: a) only a single node, b) deterministic service rate and c) at most a few hundred streams. We compare the performance of an MMPP-2 approximation to superposed IPPs for increased number of nodes (5, 7, 10), for exponential service rates and for 500 to 2500 IPP streams, each of which is better reflective of a real IP network scenario. Detailed simulation results show that the MMPP-2 approximation works exceedingly well with less than 4% average relative error and less than 6% maximum error. Furthermore, the approximation improves as the number of arriving IPPs, i.e., n increases. A subsequent example shows that this model can be utilized to calculate bandwidth bounds to ensure QoS for aggregate VoIP calls using the Expedited Forwarding class in the IP network.