Analysis and efficient provisioning of access networks with correlated and bursty arrivals

Abstract

SUMMARY The statistical multiplexer problem was analyzed extensively in the discrete‐time case within the context of ATM access networks or Internet. Despite the vast literature on ATM multiplexer, this problem has not been satisfactorily analyzed in the continuous‐time case. In this paper, a statistical multiplexer consisting of a single server queue is modeled and analyzed. N identical interrupted Poisson process arrivals flow into the multiplexer are considered, the service time is assumed to be exponentially distributed, and the queue size is infinite. We developed an efficient way to approximate the superposition of n interrupted Poisson process (or n MMPP 2) arrival process by an MMPP 2 through moments and index of dispersion count (IDC) matching. We then find the closed‐form Laplace transform of the departure process of an MMPP 2/M/1 queue. Our results prove that the departure process of an MMPP 2/M/1 queue is not an MMPP 2, which closes the long open problem that whether the departure process of an MMPP/M/1 can be a MMPP. Using Laplace transform matching, we approximated the departure process of an MMPP 2/M/1 by an MMPP 2. These results permit us to analyze a number of statistical multiplexer linked in series as they arise in an access network. Numerical experiments show that approximate analytical results match the simulation results very well. Copyright © 2013 John Wiley & Sons, Ltd.

Modelling Adversarial Flow in Software-Defined Industrial Control Networks Using a Queueing Network Model

“…Unlike them, we consider network traffic in software-defined industrial control networks. We use the results from [13] to better model the packet arrival process at the control plane. Also, we observe how adversarial flow may impact network traffic in software-defined industrial control networks.…”

“…We consider the strong periodic patterns of the network traffic in industrial control networks and approximate the arrival process using MMPP. Also, we use results from [13] to obtain a realistic characterization of the interaction between the data plane and the control plane. We then apply the analytical model to study the effect of adversarial flow in softwaredefined industrial control networks.…”

“…The arrival process of packets at the controllers in the control plane is dependent on the departure process of packets from the data plane. Tian [13] have shown that the output process of the Markovial arrival process is not a Markovial process, but rather, it becomes a Poisson process. Thus, like most existing works on the analysis of SDN using queueing network models, we employ a Poisson arrival process with exponential service time distribution (M/M/1) to model the behaviour of the controllers in the control plane [8], [10]- [12], [23]- [25].…”

Dynamic intelligence towards smart and green world

“…Despite the vast literature on ATM multiplexer, this problem has not been satisfactorily analyzed in the continuous‐time case. In the paper by Wenhong Tian entitled ‘Analysis and efficient provisioning of access networks with correlated and bursty arrivals’ , a statistical multiplexer consisting of a single‐server queue is modeled and analyzed. The authors proposed a continuous time queuing model to analyze a number of statistical multiplexer linked in series.…”

An empirical study on TCP flow interarrival time distribution for normal and anomalous traffic

“…According to their findings, confirmed by several other works, for example, [3][4][5], there are long range dependencies present in the packet interarrival time process that cannot be accurately modeled by an exponential distribution. This bursty arrival of network packets have been subject of many researches ever since (e.g., [6][7][8][9][10][11] to name just a few recent ones).…”