Cumulative Input Process Research Articles

Admission Control for Multidimensional Workload input with Heavy Tails and Fractional Ornstein-Uhlenbeck Process

The infinite source Poisson arrival model with heavy-tailed workload distributions has attracted much attention, especially in the modeling of data packet traffic in communication networks. In particular, it is well known that under suitable assumptions on the source arrival rate, the centered and scaled cumulative workload input process for the underlying processing system can be approximated by fractional Brownian motion. In many applications one is interested in the stabilization of the work inflow to the system by modifying the net input rate, using an appropriate admission control policy. In this paper we study a natural family of admission control policies which keep the associated scaled cumulative workload input asymptotically close to a prespecified linear trajectory, uniformly over time. Under such admission control policies and with natural assumptions on arrival distributions, suitably scaled and centered cumulative workload input processes are shown to converge weakly in the path space to the solution of a d-dimensional stochastic differential equation driven by a Gaussian process. It is shown that the admission control policy achieves moment stabilization in that the second moment of the solution to the stochastic differential equation (averaged over the d-stations) is bounded uniformly for all times. In one special case of control policies, as time approaches ∞, we obtain a fractional version of a stationary Ornstein-Uhlenbeck process that is driven by fractional Brownian motion with Hurst parameter H > ½.

Admission Control for Multidimensional Workload input with Heavy Tails and Fractional Ornstein-Uhlenbeck Process

The infinite source Poisson arrival model with heavy-tailed workload distributions has attracted much attention, especially in the modeling of data packet traffic in communication networks. In particular, it is well known that under suitable assumptions on the source arrival rate, the centered and scaled cumulative workload input process for the underlying processing system can be approximated by fractional Brownian motion. In many applications one is interested in the stabilization of the work inflow to the system by modifying the net input rate, using an appropriate admission control policy. In this paper we study a natural family of admission control policies which keep the associated scaled cumulative workload input asymptotically close to a prespecified linear trajectory, uniformly over time. Under such admission control policies and with natural assumptions on arrival distributions, suitably scaled and centered cumulative workload input processes are shown to converge weakly in the path space to the solution of a d-dimensional stochastic differential equation driven by a Gaussian process. It is shown that the admission control policy achieves moment stabilization in that the second moment of the solution to the stochastic differential equation (averaged over the d-stations) is bounded uniformly for all times. In one special case of control policies, as time approaches ∞, we obtain a fractional version of a stationary Ornstein-Uhlenbeck process that is driven by fractional Brownian motion with Hurst parameter H > ½.

Internet traffic modelling: from superposition to scaling

Internet traffic at various tiers of service providers is essentially a superposition or active mixture of traffic from various sources. Statistical properties of this superposition and a resulting phenomenon of scaling are important for network performance (queuing), traffic engineering (routing) and network dimensioning (bandwidth provisioning). In this article, the authors study the process of superposition and scaling jointly in a non-asymptotic framework so as to better understand the point process nature of cumulative input traffic process arriving at telecommunication devices (e.g., switches, routers). The authors further assess the scaling dynamics of the structural components (packets, flows and sessions) of the cumulative input process and their relation with superposition of point processes. Classical and new results are discussed with their applicability in access and core networks. The authors propose that renewal theory-based approximate point process models, that is, Pareto renewal process superposition and Weibull renewal process superposition can model the similar second-order scaling, as observed in traffic data of access and backbone core networks, respectively.

Poisson Shot Noise Traffic Model and Approximation of Significant Functionals

In Internet traffic modeling, many authors presented models based on particular fractal shot noise representations. The inconvenience of these approaches is the multitude of assumptions and the lack of tools to check them. In this paper we propose a unified model based on a general Poisson shot noise representation for the cumulative input process (CIP). We present a procedure of approximation of this process; then we give a procedure for controlling the bandwidth of Internet providers. The approximation and control go via limit theorems for functionals of the CIP, namely, the supremum process, the right inverse, and the storage mapping.

Modeling network traffic by a cluster Poisson input process with heavy and light-tailed file sizes

We consider a cluster Poisson model with heavy-tailed interarrival times and cluster sizes as a generalization of an infinite source Poisson model where the file sizes have a regularly varying tail distribution function or a finite second moment. One result is that this model reflects long-range dependence of teletraffic data. We show that depending on the heaviness of the file sizes, the interarrival times and the cluster sizes we have to distinguish different growths rates for the time scale of the cumulative traffic. The mean corrected cumulative input process converges to a fractional Brownian motion in the fast growth case. However, in the intermediate and the slow growth case we can have convergence to a stable Levy motion or a fractional Brownian motion as well depending on the heaviness of the underlying distributions. These results are contrary to the idea that cumulative broadband network traffic converges in the slow growth case to a stable process. Furthermore, we derive the asymptotic behavior of the cluster Poisson point process which models the arrival times of data packets and the individual input process itself.

A fluid cluster Poisson input process can look like a fractional Brownian motion even in the slow growth aggregation regime

We show that, contrary to common wisdom, the cumulative input process in a fluid queue with cluster Poisson arrivals can converge, in the slow growth regime, to a fractional Brownian motion, and not to a Lévy stable motion. This emphasizes the lack of robustness of Lévy stable motions as ‘birds-eye’ descriptions of the traffic in communication networks.

A fluid cluster Poisson input process can look like a fractional Brownian motion even in the slow growth aggregation regime

We show that, contrary to common wisdom, the cumulative input process in a fluid queue with cluster Poisson arrivals can converge, in the slow growth regime, to a fractional Brownian motion, and not to a Lévy stable motion. This emphasizes the lack of robustness of Lévy stable motions as ‘birds-eye’ descriptions of the traffic in communication networks.

Activity rates with very heavy tails

Consider a data network model in which sources begin to transmit at renewal time points { S n } . Transmissions proceed for random durations of time { T n } and transmissions are assumed to proceed at fixed rate unity. We study M ( t ) , the number of active sources at time t , a process we term the activity rate process, since M ( t ) gives the overall input rate into the network at time t . Under a variety of heavy-tailed assumptions on the inter-renewal times and the duration times, we can give results on asymptotic behavior of M ( t ) and the cumulative input process A ( t ) = ∫ 0 t M ( s ) d s .

A Sample-Path Condition for the Asymptotic Uniform Distribution of Clearing Processes

In this note we show that the asymptotic time-average distribution of a functional of cumulative input process associated with an imbedded point process follows as asymptotic uniform distribution almost surely under a mild regulatory sample-path condition. Examples from stochastic clearing processes, inventory systems and renewal theory are provided.

LIMITS FOR CUMULATIVE INPUT PROCESSES TO QUEUES

We establish functional central limit theorems (FCLTs) for a cumulative input process to a fluid queue from the superposition of independent on–off sources, where the on periods and off periods may have heavy-tailed probability distributions. Variants of these FCLTs hold for cumulative busy-time and idle-time processes associated with standard queueing models. The heavy-tailed on-period and off-period distributions can cause the limit process to have discontinuous sample paths (e.g., to be a non-Brownian stable process or more general Lévy process) even though the converging processes have continuous sample paths. Consequently, we exploit the Skorohod M1 topology on the function space D of right-continuous functions with left limits. The limits here combined with the previously established continuity of the reflection map in the M1 topology imply both heavy-traffic and non-heavy-traffic FCLTs for buffer-content processes in stochastic fluid networks.

Time-Dependent Behavior of Fluid Buffer Models with Markov Input and Constant Output Rates

The authors study a method for the analysis of fluid buffer models with Markov input rate and constant output rate. In telecommunications applications, the constant output rate typically represents a fixed-rate transmission channel. Given the initial system state, the conditional (time-dependent) survivor function and moments of the buffer content are characterized as the solution of an integral equation. The method depends on the solution of a first passage time problem for the cumulative input process (an integrated Markov process) which is described by a Fredholm integral equation of the second kind. Under appropriate conditions, the integral equation has a Neumann series solution that can be found by successive approximations. Numerical results are presented for a specific example of a buffer shared by a number of input sources varying as independent Ornstein–Uhlenbeck diffusion processes, which may be used, for instance, as a model for statistically multiplexed variable bit-rate video.

Extensions of the Queueing Relations L = λW and H = λG

This paper extends the fundamental queueing relations L = λW and H = λG that relate customer averages (the customer-average waiting time W or cost G) to associated time averages (the time-average queue length L or cost H) given an arrival process with arrival rate λ. These relations can be established by focusing on a two-dimensional cumulative input process that has the two one-dimensional cumulative input processes of interest as marginals. Relations between the marginal averages are established for cumulative input processes that may not be representable as integrals or sums. The general framework includes the continuous versions of L = λW and H = λG due to T. Rolski and S. Stidham as well as the standard version of H = λG, and can be extended to higher dimensions. Inequalities are also established when some of the conditions for equality do not hold. Moreover, central limit theorem versions of H = λG are established, extending our recent results for L = λW.

The control of a finite dam with penalty cost function: Markov input rate

The long-run average cost per unit time of operating a finite dam controlled by a policy (Lam Yeh (1985)) is determined when the cumulative input process is the integral of a Markov chain. A penalty cost which accrues continuously at a rate g(X(t)), where g is a bounded measurable function of the content, is also introduced. An example where the input rate is a two-state Markov chain is considered in detail to illustrate the computations.

The control of a finite dam with penalty cost function: Markov input rate

The long-run average cost per unit time of operating a finite dam controlled by a policy (Lam Yeh (1985)) is determined when the cumulative input process is the integral of a Markov chain. A penalty cost which accrues continuously at a rate g(X(t)), where g is a bounded measurable function of the content, is also introduced. An example where the input rate is a two-state Markov chain is considered in detail to illustrate the computations.

The control of a finite dam with penalty cost function: Wiener process input

The long-run average cost per unit time of operating a finite dam controlled by a Pτ,λM policy (Attia, 1987) is determined when the cumulative input process is a Wiener process with drift. A penalty cost which accrues continuously at the rate g(Z(t)), where g is a bounded measurable function of the content, is also introduced. We first obtain the resolvent operator Rα of a Wiener process with a reflecting boundary at 0 and the expansion of the associated kernel Kα as a power series in α. Then we use these results to determine the long-run average cost per unit time.

The control of a finite dam

The long-run average cost per unit time of operating a finite dam controlled by a PlM policy (Faddy (1974), Zuckerman (1977)) is determined when the cumulative input process is (a) a Wiener process with drift and (b) the integral of a Markov chain. It is shown how the cost for (a) can be obtained as the limit of the costs associated with a sequence of input processes of the type (b).

The control of a finite dam

The long-run average cost per unit time of operating a finite dam controlled by a Pl M policy (Faddy (1974), Zuckerman (1977)) is determined when the cumulative input process is (a) a Wiener process with drift and (b) the integral of a Markov chain. It is shown how the cost for (a) can be obtained as the limit of the costs associated with a sequence of input processes of the type (b).

Storage processes with general release rule and additive inputs

A construction is given of a process in which X(t) represents the content at time t of a dam whose cumulative input process is a Lévy process with measure v and whose release rate at time t is r(X(t)). It is assumed only that r(0) = 0 and that r is strictly positive and left-continuous with strictly positive finite right limits on (0,∞). The sample-paths of X are shown to satisfy the storage equation The process X is analyzed using renewal theory and stochastic comparison techniques, and necessary and sufficient conditions are found in terms of v and r for X to have a stationary distributionπ. These generalize previous results which were obtained under the assumption that v is finite. Conditions for Πto have an atom at 0 are considered in some detail, and related results on the positivity of the expected occupation time of level 0 are given.Necessary and sufficient conditions for the existence of Πare expressed in terms of the existence of non-negative integrable solutions of certain integral equations and conditions are given under which such solutions are necessarily stationary densities for X. A simple sufficient condition for X to have a stationary distribution is found in terms of and in the case when r is non-decreasing the condition is shown to be also necessary. Finally some examples are considered; these show that the results described above unify various known conditions in special cases, and confirm several conjectures in the related literature.

Storage processes with general release rule and additive inputs

A construction is given of a process in which X(t) represents the content at time t of a dam whose cumulative input process is a Lévy process with measure v and whose release rate at time t is r(X(t)). It is assumed only that r(0) = 0 and that r is strictly positive and left-continuous with strictly positive finite right limits on (0,∞). The sample-paths of X are shown to satisfy the storage equation The process X is analyzed using renewal theory and stochastic comparison techniques, and necessary and sufficient conditions are found in terms of v and r for X to have a stationary distributionπ. These generalize previous results which were obtained under the assumption that v is finite. Conditions for Πto have an atom at 0 are considered in some detail, and related results on the positivity of the expected occupation time of level 0 are given. Necessary and sufficient conditions for the existence of Πare expressed in terms of the existence of non-negative integrable solutions of certain integral equations and conditions are given under which such solutions are necessarily stationary densities for X. A simple sufficient condition for X to have a stationary distribution is found in terms of and in the case when r is non-decreasing the condition is shown to be also necessary. Finally some examples are considered; these show that the results described above unify various known conditions in special cases, and confirm several conjectures in the related literature.

Invariant imbedding and dams with Markovian input rate

The method of invariant imbedding, used extensively in astrophysics and in neutron transport theory, is applied to the analysis of dams whose input rate is a finite state-space continuous-parameter Markov chain. Passage-time distributions are considered as functions of the dam capacity. When the release rate is constant the method yields first-order non-linear differential equations with initial conditions obtained from the easily derived behaviour of a dam with zero capacity. Solutions of these equations are given. Some possible correlation functions for the increments of the cumulative input process are examined when the underlying Markov chain is stationary. Finally the analysis is extended to deal with dams having content-dependent release-rate of the form