V. Frost and B. Melamed, "Traffic Modeling for Telecommunications Networks", IEEE Communications Magazine, Vol. 32, No. 3, 1994, pp. 70-80.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....the most significant impact on queueing. Lazar et al. 4] developed video models for the slice and frame time scales, and showed that in the case of strict quality of service (QOS) requirements (small time delay) precise modeling of the high order autocorrelation is of secondary importance. In [5], Frost and Melamed survey a wide range of approaches to traffic modeling, several of which take multiple time scales into account, for example, self similar or fractal models. These models essentially attempt to capture an infinite number of time scales and, for that reason, they generally suffer ....

V. S. Frost and B. Melamed, "Traffic modeling for telecommunication networks," IEEE Commun. Mag., pp. 70--81, Mar. 1994.

....traffic model [12, 13, 29] Earlier traffic models are Markovian in nature, with an autocorrelation function (ACF) that drops off exponentially. Examples of these models are the autoregressive moving average (ARMA) models, Markov Arrival processes (MAP) Markov modulated processes, etc. see [10, 2, 19] for surveys of traffic models) Markovian models exhibit an exponentially decaying autocorrelation structure, which makes them short range dependent (SRD) An SRD model is one for which the ACF is summable, i.e. k ae k 1. Note, however, that an SRD model is not necessarily Markovian. In ....

.... 500,000 points (the computational complexity involved in generating M=G=1 traces grows linearly with the trace length, while this complexity grows quadratically in the case of F ARIMA traces) Figure 12 indicates asymptotic slopes of Gamma0:79 and Gamma0:75 for aggregation levels in the ranges [10 ] and [10 ] respectively. This would suggest that the underlying data exhibit LRD. However, we know that the data were generated from an SRD M=G=1 model Despite the length of the M=G=1 trace, the VT test may wrongly suggest the presence of LRD in this trace (if one is not careful in ....

[Article contains additional citation context not shown here]

Victor S. Frost and Benjamin Melamed. Traffic modeling for telecommunications networks. IEEE Communications Magazine, 32(3):70--81, March 1994.

....in the remainder of the contribution. The material is provided for information, and has, in general, not been discussed in T1X1.3. A general treatment of these topics at a basic level is given in Volume 1 of [5] A more advanced treatment is given in [6] A survey of traffic models is given in [7]. 2.1 Types of models In this contribution, we refer to continuous time, discrete time, and discrete event models (see [8] for a more complete discussion) In a continuous time model, the state can change at any instant of time. Usually the state changes are continuous. In a discrete time ....

Victor S. Frost and Benjamin Melamed, Traffic Modeling for Telecommunications Networks, IEEE Communications Magazine, Vol. 32, No. 3, March 1994, pp. 70-81.

....synchronized to frame starts. All data is discretized into ATM cells carrying 48 octets of payload, i.e. the frame sizes, 3 the buffer content, and the amountofdatatransmitted during a frame duration. Cells of a single frame are regarded as a fluid according to the fluid simulation approach [2]. Instead of individual cell arrivals we consider frame data as a fluid, which flows into the buffer at a constant rate. There are two important benefits from the fluid approach. It is conceptually simple and it leads to a reduced algorithmic complexity while the accuracy of the results is ....

V. S. Frost and B. Melamed. Traffic modeling for telecommunication networks. IEEE Communications Magazine, 32(3):70--81, Mar. 1994.

....of active set of users is possible with a simple matched filterbank at the first stage. I. MOTIVATION AND INTRODUCTION I N packet switching networks, it has been observed that the conventional Poisson assumption is not adequate to capture the traffic variability and dependence (see, e.g. [5]) In particular, traffic burstiness the fact that packets usually come in bursts has been revealed and carefully studied in the past decades. From the macroscopic point of view, traffic burstiness results in unpredictability of the overall network traffic [11] However, from a microscopic ....

V. Frost and B. Melamed, "Traffic modeling for telecommunications networks, " IEEE Commun. Mag., Mar. 1994.

....MPEG video, but its correlation function still contains a single real exponential as a two state Markov chain. In [14] Jagerman and Melamed proposed a measurement based modeling technique called transform expand sample (TES) which is mainly suitable for computer simulation. In most TES examples [15], only consists of one or two exponentials. Recent traf fic measurement [1, 2] also identifies the significance of long range dependencies, which are described by the correlation behavior in large time scales (or, equivalently by the dominant power spectrum in low frequency band) The study ....

V. S. 1%ost, B. Melamed, "Traffic Modeling for Telecom- munications Networks," IEEE Communications Magazine, pp. 70-81, March 1994.

.... called transform expand sample (TES) A TES process matches through an inverse transformation of a background process to the foreground process The matching of is achieved by a visual interactive tuning of the parameters which determines This technique has been applied to match the video traces [3], 17] 21] Recent traffic measurement [11] also identifies the significance of long range dependencies, which are described by the correlation behavior in large time scales (or, equivalently, by the dominant power spectrum in low frequency band) In fact, 12] 16] and [7] indicate that a ....

V. S. Frost and B. Melamed, "Traffic modeling for telecommunications networks," IEEE Commun. Mag., pp. 70--81, Mar. 1994.

....a traffic model [10] 11] 23] Earlier traffic models are Markovian in nature, with an autocorrelation function (ACF) that drops off exponentially. Examples of these models are the autoregressive moving average (ARMA) models, Markov Arrival processes (MAP) Markov modulated processes, etc. see [8], 1] 17] for surveys of traffic models) Markovian models exhibit an exponentially decaying autocorrelation structure, which makes them short range dependent (SRD) An SRD model is one for which the ACF is summable, i.e. k ae k 1. Note, however, that an SRD model is not necessarily ....

Victor S. Frost and Benjamin Melamed. Traffic modeling for telecommunications networks. IEEE Communications Magazine, 32(3):70--81, March 1994.

....this fact does not hold in presence of LRD. 1 Introduction Historically, traffic modeling has its origins in the conventional telephony, and it has been based almost exclusively on supposing independent interarrival times (fundamentally Poisson models) and exponentially distributed service times [11]. The fundamental reason for these suppositions is to be able to get relatively simple models from the analytical point of view. However, multimedia traffic, present in the highspeed networks, is characterized by high variability (burstiness) and strong positive correlation [10, 22] much more ....

V. Frost and B. Melamed. Traffic modeling for telecommunications networks. IEEE Communications Magazine, 32:70--80, 1994.

....size, spatial complexity, and motion were estimated as defined in Equations 2.4, 2.5, and 2.6 respectively. 2. 4 Traffic within Activity Periods Despite a large number of VBR models proposed over the last decade, no model appears to be suitable for all different types of real world video traffic [30]. VBR video exhibits complex characteristics (e.g. self similarity, variable time scale property, long term dependency and non stationarity) that make model identification and estimation very hard. In general, there is a tradeoff between model complexity and accuracy. To simplify model ....

....used for the real time traffic prediction in dynamic resource allocation. Encoder specific synthetic traffic generators are used as a traffic source for simulation of buffering requirements and cell loss estimation at the network multiplexer. A good overview of traffic modeling can be found in [30]. Typically, traffic models are based on renewal processes (special case Poisson and Bernoulli) Markov renewal processes, Markov processes, Markov modulated processes (special case Markovmodulated Poisson Process, fluid traffic models) and autoregressive models (AR, MA, ARMA, ARIMA) 101, 102] ....

V. S. Frost and B. Melamed, "Traffic Modeling For Telecommunications Networks ", IEEE Communications Magazine, March 1994, pp. 70-81.

....a two state Markov Chain model based on a statistical analysis of 16 experimental telephone conversations. Figure1 illustrates the talkspurt state, in which a voice source generates signals in audible frequencies, and silence state, in which the generated signals are in an inaudible range[4]. The proposed On Off (or talkspurt silence) model builds the basis of a variety of later research work on the specification of voice sources, see Figure 2. In the 80 s and early 90 s, although the significance of the effect of traffic correlation on queueing performance was recoganized, early ....

....details within the OPNET node model. Figure 3 shows the node model. Fig. 3 On Off Node The node includes 2 modules: a process (on off source) and a transmitter (pt 0) A packet stream is used to connect the two modules, which represent the flow of data across within a communication node[4].Transmitter modules serve as the interface between packet streams inside a node and communication links outside the node. Here we use a point to point transmitter. The finite state machine mechanism is used inside the process module. Figure 4 shows the state transmission graph of the ....

V. S. Frost and Benjamin Melamed, "Traffic Modeling for Telecommunications Networks", IEEE Communication Magazine, Vol.32,No 3, pp. 70-81, 1994.

.... decade of teletraffic research [8,11,14,19,22] Source models of VBR video are needed to dimension networks and control methods to achieve acceptable quality and optimal usage of network resources [4,8] A number of different models have been proposed for VBR video modeling (for an overview see [9,25]) The large variety of modeling approaches can be divided into the following three categories: Markov models [1,17] autoregressive processes [10,23] and fractal models [18,28] However, all common to these approaches that a specific stochastic process is chosen and the parameters are set by a ....

V. S. Frost and B. Melamed, Traffic Modeling For Telecommunications Networks, IEEE Communications Magazine, (March 1994)70-81. 35

....summable correlation structures which typically arise in Markovian traffic models and Poisson like sources. This state of affairs has generated a strong interest in a number of alternative traffic models which capture observed (long range) dependencies; good surveys are available in [13], 22] 26] Proposed models include Fractional Brownian Motion [25] and its discrete time analog, Fractional Gaussian Noise [1] on off sources with subexponential activity periods [15] and references therein) and traffic model with subexponential session duration [28] Under these new ....

V. S. Frost and B. Melamed, "Traffic modeling for telecommunications networks," IEEE Communications Magazine 32 (1994), pp. 70--81.

....such characteristics: i) the marginal distribution, ii) the autocorrelation structure and (iii) the correspondence between the sample paths of the simulated and the empirical sequence. The recently developed TES (Transform Expand Sample) method differs from traditional modeling techniques [1] in so far as it claims to strive with all three of these objectives simultaneously [2] In fact, it successfully reproduces the exact marginal distribution and approximates the autocorrelation function sufficiently well, but this objective as well as the aim of rendering sample paths that look ....

Frost, V.; Melamed, B.: Traffic Modeling for Telecommunications Networks. IEEE Communications Magazine (March 1994), 70 -- 81.

....they can be used to generate upper bounds on loss and delay of traffic. These models are called bounded traffic models. Michiel, 1997] Both stochastic models and bounded models are used in developing our simulation models. One popular traffic model is the Poisson model [Schwartz, 1987] [Frost, 1994]. This model has been used since the telephone era, where it was effective at modeling the times at which telephone calls arrived at a switch. A Poisson process is a memoryless, independent and identical (i.i.d) process. The interarrival times are exponentially distributed and the number of ....

....number of independent traffic sources. The theoretical basis for this phenomenon is known as Palm s Theorem [Larson, 1979] It roughly states that under suitable but mild conditions, such multiplexed streams, the statistics of the sum approaches a Poisson process as the number of streams grows. [Frost, 1994] Thus, traffic streams on main communications trunks are commonly believed to follow Poisson arrival statistics, as opposed to traffic on upstream links, which is less likely to be Poisson. The Poisson model is simple and can be used with care for simulating the background traffic of a network, ....

Victor S. Frost and Benjamin Melamed, "Traffic Modeling for Telecommunications Networks", IEEE Communication Magazine, March 1994

....insure that they do not exceed their negotiated traffic characteristic parameters. Performance modeling techniques are needed to determine which congestion control techniques should be used. Performance modeling techniques include: analytical techniques, computer simulation, and experimentation [1]. Performance models require accurate traffic models which can capture the statistical characteristics of actual traffic. If the traffic models do not accurately represent actual traffic, one may overestimate or underestimate network performance. This article surveys traffic models in ....

....chain, and is referred to as an embedded Markov chain. In a simple Markov traffic model, each state transition represents a new arrival. Therefore, inter arrival times are exponentially distributed (for continuous time case) and their rates depend on the state from which the transition occur [1]. The rest of this section discusses various Markov and embedded Markov models that have been used to model network traffic. ON OFF AND IPP MODELS The ON OFF source model shown in Fig. 2a is the most popular source model for voice [4, 5] In this model, packets are only generated during talk ....

[Article contains additional citation context not shown here]

V. Frost and B. Melamed, "Traffic Modeling for Telecommunications Networks," IEEE Commun. Mag., Mar. 1994.

.... The call traffic model describes how often individual users place calls to other users and characterizes the duration of each call. Very little is known about the traffic characteristics of future PCS networks. However, on fixed telephone networks, traffic is modeled accurately. Reference [17] gives an overview of existing call traffic models. For current telephone usage, according to [6] the mean call arrival rate and mean call duration during busy hours are 2.8 calls hour and 2.6 min call, respectively. Our call traffic model generates call arrivals (i.e. calls initiated) for ....

....call arrival rates of users over the reference time periods and their respective call and relative call probabilities. MOBILITY MODELS Before discussing the mobility models we developed, we first review some common approaches for modeling human movements. Fluid Model The fluid model [6, 17, 18] conceptualizes traffic flow as the flow of a fluid. It is used to model macroscopic movement behavior. In its simplest form, the model formulates the amount of traffic flowing out of a region to be proportional to the population density within the region, the average velocity, and the ....

V. S. Frost and B. Melamed, "Traffic Modeling for Telecommunications Networks," IEEE Commun. Mag., vol. 32, no. 3, Mar. 1994, pp. 70--81.

....reproduce the characteristic features of a given empirical time series as closely as possible. Such characteristics may be (i) the marginal distribution, ii) the autocorrelation structure or simply (iii) the visual appearance of the empirical sequence. Traditional methods for traffic modeling [1] use to concentrate on one or two of the mentioned three objectives, whereas the recently developped TES (Transform ExpandSample) method claims to strive for coping with all three objectives simultaneously [2] TES is capable of reproducing the marginal distribution exactly and the autocorrelation ....

Frost, V.; Melamed, B.: Traffic Modeling for Telecommunications Networks. IEEE Communications Magazine (March 1994), 70 -- 81.

....Model The call traffic model describes how often individual users place calls to other users and characterizes the duration of each call. Very little is known about the traffic characteristics of future PCS networks. However, on fixed telephone networks, traffic is modeled accurately. Reference [8] gives an overview of existing call traffic models. For current telephone usage, according to [21] the mean call arrival rate and the mean call duration during busy hours are 2.8 calls hour and 2.6 min call, respectively. Our call traffic 3 model generates call arrivals (i.e. calls initiated) ....

....average call arrival rates of users over the reference time periods and their respective call and relative call probabilities. 4.2 Mobility Models Before discussing the mobility models we developed, we first review some common approaches for modeling human movements. Fluid Model The fluid model [8, 19, 21] conceptualizes traffic flow as the flow of a fluid. It is used to 6 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 Relative Call Probability (Rank 2) Month Week Day Cumul. Prob. of Occurrence Relative Call Probability (Rank 3) 0 0.2 0.4 0.6 0.8 1 Month Week Day 0 0.2 0.4 0.6 ....

V.S. Frost and B. Melamed. Traffic modeling for telecommunications networks. IEEE Communications Magazine, 32(3):70--81, March 1994.

....the behaviour of the candidate burstiness descriptors which could take us a step closer to establishing a generally acceptable burstiness measure. 1 Introduction Burstiness has been found a key characteristic of broadband traffic which plays a critical role in determining network performance [5, 18]. However, the nature of bursty traffic is still poorly understood, moreover, there is not even a single and widely accepted notion of burstiness in the teletraffic literature. Burstiness expresses the clustering phenomenon of arrivals, that is, when arrivals tend to form clusters with relatively ....

....characteristics of the marginal distribution of the interarrival time. A set of candidates are the moments of that distribution. However, in practice the peak to mean ratio and the squared coefficient of variation are the most frequently used first order measures in the teletraffic literature [5, 18]. Measures expressing second order properties of the traffic are more complex. The indices of dispersion [7, 20] and the generalized peakedness [2, 3] are the most well known measures from this class. The indices of dispersion measures include the correlation properties of the traffic and can be ....

[Article contains additional citation context not shown here]

V. S. Frost, B. Melamed, "Traffic Modeling For Telecommunications Networks", IEEE Communications Magazine, March, 1994.

....traffics. NTCD MB can satisfy both the delay and the loss rate of real time traffics, but it does not meet the different levels of loss rate requirements for the multiple classes of nonreal time traffics. 2.2. Traffic Model The traffic models can be classified into the following six categories [10]. 2.2.1.Renewal traffic models In these models the cell interarrival time is assumed to have independent identical distribution for each individual traffic. Although theirs mathematical computation is simple but the superposition of independent renewal process does not produce a renewal process, ....

....mathematical computation is simple but the superposition of independent renewal process does not produce a renewal process, with few exceptions. Also it does not capture the autocorrelation so it cannot explain traffic burstiness, which is the major characteristic of the broadband network an LAN [8,10,11]. 2.2.2.Markov and Markov renewal traffic models The Markov and Markov renewal traffic models introduce dependency into the random sequence of interarrival time and they can capture traffic burstiness. In a Markov traffic model interarrival time are exponentially distributed and their rate ....

[Article contains additional citation context not shown here]

V.S.Frost and B.Melamed "Traffic modeling for telecommunications networks," IEEE Comm. Mag., Mar. 1994 PP. 70-81.

....of the behaviour of the candidate burstiness descriptors which could take us a step closer to establishing a generally acceptable burstiness measure. 1 Introduction Burstiness has been found a key characteristic of broadband traffic which plays a critical role in determining network performance [4, 15]. However, the nature of bursty traffic is still poorly understood, moreover, there is not even a single and widely accepted notion of burstiness in the teletraffic literature. Burstiness expresses the clustering phenomenon of arrivals, that is, when arrivals tend to form clusters with relatively ....

....characteristics of the marginal distribution of the inter arrival time. A set of candidates are the moments of that distribution. However, in practice the peak to mean ratio and the squared coefficient of variation are the most frequently used first order measures in the teletraffic literature [4, 15]. Measures expressing second order properties of the traffic are more complex. The indices of dispersion [6, 17] and the generalized peakedness [2] are the most well known measures from this class. The indices of dispersion measures includes the correlation properties of the traffic and can be ....

[Article contains additional citation context not shown here]

V. S. Frost, B. Melamed, "Traffic Modeling For Telecommunications Networks", IEEE Communications Magazine, March, 1994.

....of the VBR traffic will be crucial to many important applications such as controlling the Quality of Service, effectively allocating network resources and designing buffer capacity of networks. Numerous studies have been conducted on traffic modeling and performance analysis, see for example [9][6] 12] 11] 21] and references therein. One of the significant statistical properties of VBR video traffic has been found to be the co existence of the socalled long range dependence (LRD) and the short range dependence (SRD) in the video trace [2] 10] 6] Roughly speaking, this means that the ....

V. Frost and B. Melamed. Traffic modeling for telecommunications networks. IEEE Communication Magazines, 32, 1994.

....diverse statistical characteristics and Quality of Service (QoS) requirements, modeling different types of traffic, and generating synthesized traffic from a model will be crucial to network design and network simulation. In this work, we will focus on these important issues. Two of the key issues [14] of accurate traffic modeling are performance and computational efficiency. The former addresses the ability of a model to characterize significant statistical properties in network traffic. The latter deals with the complexity of a model, and the computational complexity needed to develop such a ....

....needed to develop such a model and to generate a large volume of synthesized traffic. Both issues are important for simulating high speed networks with extremely low loss probability. Although numerous studies have been conducted on traffic modeling and performance analysis (see for example [14][33] 38] 43] 41] and references herein) new traffic models are needed to deal with new properties introduced by data and video traffic. One of such significant statistical properties is the so called long range dependence (LRD) recently found in Ethernet data[21] and VBR video traffic[15] It is ....

[Article contains additional citation context not shown here]

V. Frost and B. Melamed. Traffic modeling for telecommunications networks. IEEE Communication Magazines, 32:70--80, 1994.

....the frame sizes of MPEG encoded video sequences and on the queueing performance at an ATM multiplexer when video frame sequences are transmitted through an ATM network. Video traffic modeling has been studied extensively and many results have been reported (for an overview, see, e.g. 1] [8], and [24] The modeling approach for VBR video traffic can roughly be divided into several main classes, i.e. histogram based models (e.g. 31] 32] Markov chain models (e.g. 13] 14] 17] 22] 27] and [29] AR processes (e.g. 5] 11] 16] 19] 20] 22] 25] 30] 33] ....

V. S. Frost and B. Melamed, "Traffic modeling for telecommunications networks," IEEE Commun. Mag., vol. 32, pp. 70--81, Mar. 1994.

....can be very valuable as they may allow predictions of performance over a wide range of scenarios. However, deriving such analytic models and approximations can be very difficult and so far there has been little experience in building models for important patterns such as selfsimilar traffic [ Frost and Melamed, 1994; Garrett and Willinger, 1994] Development of analytic models requires extensive effort to validate the models against actual measured data. Another approach to understanding and predicting performance is to use detailed simulations of the traffic. Such simulations can also be used to validate ....

Frost, V., and Melamed, B.(1994) "Traffic Modeling for Telecommunications Networks," IEEE Communications, 32(3), pp. 70-81, March.

....fidelity traffic models as the outputs from a simulation (used for network performance estimation) are highly dependent on the inputs provided to the model. Hence without realistic traffic source models, the simulation results are of little or no value (Arlitt, Chen, Gurski and Williamson, 1995) (Frost and Melamed, 1994). Classical techniques used to model traffic in previous networks have not been found to hold for broadband networks such as ATM. The most striking reason behind this effect is that traffic sources for packetised data flows reveal bursty and correlated statistical behavior affecting system ....

....in (Cleary, Pearson, Graham Unger, 1996) Another issue that needs to be addressed when designing traffic models is the detail that is going to be modeled for a particular traffic source. The designer of a traffic model must be aware of the trade off between detail and simulation execution time (Frost and Melamed, 1994). For example, consider the simulation of an ATM network to estimate the cell loss ratio (CLR) which has the acceptable value of 10 7 . In order to have confidence in network dimensions, significant measures of CLR need to be made requiring simulation of between 10 9 and 10 10 cells. Even ....

[Article contains additional citation context not shown here]

Frost, V. and Melamed, B. "Traffic Modeling for Telecommunications Networks", IEEE Communications, Vol. 32, No. 3, pp. 70-81, March 1994.

.... and Summary Traffic measurements and traffic models have long played an important role in telecommunications systems, e.g. see Cole [5] and Rahko [17] However, the emergence of new services and new technologies has led to new kinds of traffic and new traffic models, e.g. see Frost and Melamed [8], Leland, Taqqu, Willinger and Wilson [1] Meier Hellstern, Wirth, Yan and Hoeflin [15] and Part I of Roberts [18] Just as with overflow traffic associated with alternative routing, in many new situations the classical Poisson process traffic model is not nearly appropriate. On the other hand, ....

V. Frost and B. Melamed, Traffic modeling for telecommunication networks, IEEE Communications Magazine 32 (1994) 70-81.

....design to data compression and representation. Network design can benefit from advanced signal processing in many ways that have yet been explored. For example, efficient network resource allocation requires estimation and prediction of traffic patterns. The modeling of multimedia network traffic [78], 3] has already attracted re MARCH 1999 IEEE SIGNAL PROCESSING MAGAZINE 43 search attention in the signal processing community [162] 1] The context of the signal processing problems in networks is different, but the fundamentals are the same. There is no doubt that signal processing will ....

V.S. Frost and B. Melarned, "Traffic modeling for telecommunications networks," IEEE Commun. Mag., vol. 4, no. 3, pp. 70-81, Mar. 1994.

....is scene length, c is scene complexity and is scene motion. Detailed discussions about particular descriptor parameters are included in section IV. We are investigating an innovative approach that scene complexity and motion could be approximated by using random walk or auto regressive AR model [2]. This approach comes naturally from the definition of a scene and its high temporal frame correlation. The changes in the bit rate during the scene are due to object complexity changes, object movement, or due to camera operations (affecting all objects) Each frame in the scene sequence can be ....

....for complexity and motion respectively. In terms of selecting the actual stochastic process for R and M, we have chosen the random walk for its relative low computation requirements and high correlation between close samples. More complex models could be chosen, such as AR, DAR or TES if necessary [2, 3, 4, 5, 6]. The following were selected as Table 1: x s 2 max x i x Low 0.6632 0.1818 3.6494 Medium 1.0870 0.4001 3.5083 High 3.5176 2.1060 2.5445 50 100 150 200 250 time [s] scalable source 1 2 3 4 5 6 7 rate [Mbps] h m l l low layer m medium layer h high layer Figure 3. ....

V. S. Frost and B. Melamed, "Traffic Modeling For Telecommunications Networks", IEEE Communications Magazine, March 1994

....in real time for live video as well. 3. 2 Scene Resource Model Despite a very large number of VBR video traffic models proposed over the last couple years, no model was flexible enough to be effectively used in different applications in which various types of real video traffic need to be modeled [8]. It appears that the selection of the appropriate traffic model, the structure and the complexity is influenced by the intended application. Several models directly applicable to dynamic resource allocation have been recently proposed. Besides the peak rate and effective bandwidth allocation ....

V. S. Frost and B. Melamed, Traffic Modeling For Telecommunications Networks, IEEE Communications Magazine, March 1994, pp. 70-81.

....the most significant impact on queueing. Lazar et al. 76] developed video models for the slice and frame time scales, and showed that in the case of strict quality of service (QOS) requirements (small time delay) precise modeling of the high order autocorrelation is of secondary importance. In [48], Frost and Melamed survey a wide range of approaches to traffic modeling, several of which take multiple time scales into account, for example, self similar or fractal models. These models essentially attempt to capture an infinite number of time scales, and for that reason, they generally suffer ....

V. S. Frost and B. Melamed. Traffic modeling for telecommunication networks. IEEE Communications Magazine, pages 70--81, March 1994.

....INTRODUCTION Over the last few years it become clear that the video component of the emerging multimedia technology will play an important role in future communication and storage systems. This underlines the key importance of video traffic modeling in the development of future multimedia systems [1]. Generally, the VBR video rate depends on video style, content and compression technique used. For example, the different video styles (videophone, movie, news, sport, etc. can have distinct statistics of scene length, etc. Also, the captured video content, such as scene background, static or ....

V. S. Frost and B. Melamed, "Traffic Modeling For Telecommunications Networks", IEEE Communications Magazine, March 1994

....a constant number of requests) or as a transaction load (giving a (stochastic) arrival rate of requests) Characterizing the arrival behavior as a stochastic process is a well known technique and does not have to be explained in detail here. The interested reader is referred to [12] or [15] or [9]. Note, that the user behavior is typically represented at the application layer [11] i.e. user submits a whole program, and not just its components. But in many scheduling or mapping studies, the objects that can be scheduled or mapped are at a finer granularity. In this case, the arrival of ....

V. S. Frost and B. Melamed. Traffic modeling for telecommunications networks. IEEE Communications Magazine, pages 70--80, March 1994.

....the previous results in [1] to an end to end rate distortion analysis of the system, we will need to characterize video sources with reasonably accurate and analytically tractable traffic models. There has been an enormous amount of interest and research in traffic modeling of VBR video sources [14] [16] In this work, we make use of the simple 2 state MMPP, illustrated in Fig. 5, whose parameters are fitted to traces produced by an MPEG 2 encoder. The MMPP, a non renewal, doublystochastic Poisson process whose arrival rate depends on the state of an underlying continuous time Markov chain, ....

V. Frost and B. Melamed, "Traffic modeling for telecommunications networks, " IEEE Communications Magazine, pp. 70--81, March 1994.

....we represent the arrival process by an N state Markov process, where N 2. Of particular interest here is the case of N = 2 as it represents the common on off behavior of network traffic. Markovian processes have been extensively used in wired networks to characterize various types of traffic (see [15] for details) While other, non Markovian classes of models, including self similar (e.g. 16, 12, 17] and subexponential models (e.g. 18] have also been proposed as a means of capturing the persistent correlations in network traffic, the jury is still out on whether such models provide ....

Victor S. Frost and Benjamin Melamed, "Traffic modeling for telecommunications networks," IEEE Communications Magazine, vol. 32, no. 3, pp. 70--81, Mar. 1994.

....on the scene activity and the type of compression involved. Hence, the output of a video codec is a VBR stream. Unlike conventional data networks, ATM networks provide efficient support for VBR traffic. Several traffic models have been proposed to characterize compressed video streams (see [1] for a survey) The parameters of these models were obtained by matching certain statistical characteristics of an actual video sequence and the model under consideration. Particular emphasis was on matching the mean, variance, and more importantly the correlation structure of the bits per frame ....

V. S. Frost and B. Melamed, "Traffic modeling for telecommunications networks," IEEE Communications Magazine, vol. 32, pp. 70--81, Mar. 1994.

....events. If the communication events in a traffic flow arrive at times t 1 ; t 2 ; t j , the difference between the arrival times of any two neighbor events is a random variable, denoted as x, called the interarrival time. Assume that the local user traffic flow is a Poisson process [6]. So its interarrival times are exponentially distributed. Its probability density function (pdf) is defined as: f(x) ffe Gammaffx , where 1 ff is the mean. The number of arrivals over a given interval (t; t x) has a Poisson distribution. Each workstation may inject several local user ....

V. Frost and B. Melamed, "Traffic Modeling for Telecommunications Networks", IEEE Communications Magazine, Vol.32, No.3, pp.70-80, 1994.

....ATM traffic 1 model by ATM simulations. For example, in ATM TN which is a CANARIE project it is one of the three traffic models. When modeling Ethernet traffic, packet and connection arrivals are often assumed to be Poisson processes because such processes have attractive theoretical properties [1]. A number of studies have shown, however, that for both local area and wide area traffic, Poisson processes are valid only for modeling the arrival of user sessions (TELNET connection and FTP control connection) that they fail as accurate model for other LAN and WAN arrival processes, and that ....

V. Frost and B. Melamed. Traffic modeling for telecommunications networks. IEEE Communications Magazine, 32(3):70--80, March 1994.

....to form visual clusters on the time line. The mathematical underpinningsof burstiness are complex; two main factors contributing to traffic burstiness are the marginal distribution and autocorrelation function of traffic interarrival times, particularly strong positive short term autocorrelations [3]. The impact of autocorrelation in traffic processes on queueing measures such as mean queue length, mean waiting times and loss probabilities in finite buffers, can be very dramatic, even in light traffic [6, 10, 14] The studies cited above support the view that modeling a bursty traffic stream ....

.... source modeling has been confined to short sequences of empirical records consisting of one or a few scenes; furthermore, aditional video source models have been designed to capture either the distribution function or the autocorrelation function of empirical bit rate records, but not both; see [3, 4] and references therein. Since traffic burstiness can be caused by the shape of the bit rate distribution as well as by its autocorrelation function, it is important to faithfully capture both first order and second order statistics of the empirical data. 120 This paper presents a novel approach ....

Frost, V. and Melamed, B. "Traffic Modeling for Telecommunications Networks", Communications Magazine, 1994 (to appear).

No context found.

V. Frost and B. Melamed, "Traffic Modeling for Telecommunications Networks", IEEE Communications Magazine, Vol. 32, No. 3, 1994, pp. 70-80.

No context found.

Frost, V.; Melamed, B.: Traffic Modeling for Telecommunications Networks. IEEE Communications Magazine (March 1994), 70 -- 81.

No context found.

Victor S. Frost and Benjamin Melamed. Traffic Modeling For Telecommunications Networks. IEEE Communications Magazine, March 1994. (p 37)

No context found.

V. Frost and B. Melamed, "Traffic Modeling for Telecommunications Networks," IEEE Communications Magazine, vol. 32, no. 3, pp. 70-81, March 1994.

No context found.

V.S. Frost and B. Melamed. Traffic Modeling for Telecommunications Networks. In IEEE Communications Magazine, March 1994.

No context found.

Frost, V.; Melamed, B.: Traffic Modeling for Telecommunications Networks. IEEE Communications Magazine (March 1994), 70 -- 81.

No context found.

V. S. Frost and B. Melamed. Traffic modeling for telecommunication networks. IEEE Communications Magazine, pages 70--81, Mar. 1994.

No context found.

Frost, Victor S. and Melamed, Benjamin, "Traffic Modeling For Telecommunications Networks," IEEE Communications Magazine, March 1994, pp. 70-81.

No context found.

V. S. Frost and B. Melamed, "Traffic modeling for telecommunications networks," IEEE Com- munications Magazine, vol. 32, pp. 70-81, March 1994.

No context found.

V. S. Frost and B. Melamed, "Traffic Modeling For Telecommunications Networks", IEEE Communications, vol.32, no. 4, Mar.1994, pp. 70 - 81.

First 50 documents  Next 50

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC