Large deviation properties of a correlated arrival process

Large Deviations of Realized Volatility

This paper deals with the stationary analysis of the finite, single server queue in discrete time. The following stntionary distributions and other quantities of practical interest are investigated: (1) the joint density of the queue length and the residual service time, (2) the queue length distribution and its mean, (3) the distribution of the residual service time and its mean, (4) the distribution and the expected value of the number of customers lost per unit of time due to saturation of the waiting capacity, (5) the distribution and the mean of the waiting time, (6) the asymptotic distribution of the queue length following departuresThe latter distribution is particularly noteworthy, in view of the substantial difference which exists, in general, between the distributions of the queue lengths at arbitrary points of time and those immediately following departures.

This paper investigates the distribution of delay and peak age of information in a communication system where packets, generated according to an independent and identically distributed Bernoulli process, are placed in a single-server queue with first-come first-served discipline and transmitted over an additive white Gaussian noise (AWGN) channel. When a packet is correctly decoded, the sender receives an instantaneous error-free positive acknowledgment, upon which it removes the packet from the buffer. In the case of negative acknowledgment, the packet is retransmitted. By leveraging finite-blocklength results for the AWGN channel, we characterize the delay violation and the peak-age violation probability without resorting to approximations based on large deviation theory as in previous literature. Our analysis reveals that there exists an optimum blocklength that minimizes the delay violation and the peak-age violation probabilities. We also show that one can find two blocklength values that result in very similar average delay but significantly different delay violation probabilities. This highlights the importance of focusing on violation probabilities rather than on averages.

We develop a strongly efficient rare-event simulation algorithm for computing the tail of the steady-state waiting time in a single server queue with regularly varying service times. Our algorithm is based on a state-dependent importance sampling strategy that is constructed so as to be straightforward to implement. The construction of the algorithm and its asymptotic optimality rely on a Lyapunov-type inequality that is used to bound the second moment of the estimator. The solution to the Lyapunov inequality is constructed using fluid heuristics. Our approach takes advantage of the regenerative ratio formula for the steady-state distribution--and does not use the first passage time representation that is particular to the delay in the G/G/1 queue. Hence, the strategy has the potential to be applied in more general queueing models.

[1] We used the frameworks of queuing theory and connectivity to study the runoff generated under constant rainfall on a one-dimensional slope with randomly distributed infiltrability. The equivalence between the stationary runoff-runon equation and the customers waiting time in a single server queue provides a theoretical link between the statistical description of infiltrability and that of runoff flow rate. Five distributions of infiltrability, representing soil heterogeneities at different scales, are considered: four uncorrelated (exponential, bimodal, lognormal, uniform) and one autocorrelated (lognormal, with or without a nugget). The existing theoretical results are adapted to the hydrological framework for the exponential case, and new theoretical developments are proposed for the bimodal law. Numerical simulations validate these results and improve our understanding of runoff-runon for all of the distributions. The quantities describing runoff generation (runoff one-point statistics) and its organization into patterns (patterns statistics and connectivity) are studied as functions of rainfall rate. The variables describing the wet areas are also compared to those describing the rainfall excess areas, i.e., the areas where rainfall exceeds infiltrability. Preliminary results concerning the structural and functional connectivity functions are provided, as well as a discussion about the origin of scale effects in such a system. We suggest that the upslope no-flow boundary condition may be responsible for the dependence of the runoff coefficient on the scale of observation. Queuing theory appears to be a promising framework for runoff-runon modeling and hydrological connectivity problems.

We consider a single server queue in discrete time, in which customers must be served before some limit sojourn time of geometrical distribution. A customer who is not served before this limit leaves the sys- tem. The fact of serving customers, holding them in queue or losing them induce costs. The purpose is to decide when to serve the customers so as to minimize these costs. We use a Markov Decision Process with infinite horizon and discounted criterion. We establish the structural properties of the stochastic dynamic programming operator, and we deduce that the optimal policy is of threshold type, and we compute the threshold explicitly.

Self-similar (or fractal) stochastic processes were proposed as more accurate models of certain categories of traffic (e.g., Ethernet traffic, variable-bit-rate video) which will be transported in ATM networks. Existing analytical results for the tail distribution of the waiting time in a single server queue based on fractional Gaussian noise and large deviation theory, are valid under a steady-state regime and for an asymptotically large buffer size. However, the predicted performance based on steady-state regimes may be overly pessimistic for practical applications. Theoretical approaches used to obtain the transient queueing behavior and queueing distributions for a small buffer size become quickly intractable. The approach we followed was based on fast simulation techniques for the study of certain rare events such as cell losses with very small probability of occurrence. Our simulation experiments provide an insight on the transient behavior that is not possible to predict using current analytical results. Finally they show good agreement with existing results when approaching steady-state.

We consider single-server queue, which models a telecommunication node. Packets arrive in the system according to the Markovian arrival process (MAP). The packets bring some amount of information, which is stored in a continuous-type unlimited buffer before it will be transmitted out of the system. We consider the system with fluid manner of service, which means that the server takes the information from the buffer with constant speed. Our aim is to derive the probability distribution of the real waiting time in the system.

Conventional wisdom suggests that a fast single-server queue is preferable to a number of slower single-server queues. The rationale is that if the service rate of the superposed queue is the sum of the service rates of the individual queues, then the average waiting time for the superposed queue is smaller than the weighted average waiting time of the individual queues. This argument, however, assumes that the service time of the customers is perfectly scalable. Unfortunately, this assumption rarely holds in reality. In this paper we analyze the superposition of M/G/1 queues. We assume that there are scalable and unscalable components in the service time. On the basis of the criteria of average waiting time and average response time, we derive necessary and sufficient conditions for whether the superposition of queues is preferable. Examples of magnetic rotational disks are given to illustrate our findings.

The stationary waiting time and other variables in single-server queues with specialities at the beginning of a busy period

Conventional wisdom suggests that a fast single-server queue is preferable to a number of slower single-server queues. The rationale is that if the service rate of the superposed queue is the sum of the service rates of the individual queues, then the average waiting time for the superposed queue is smaller than the weighted average waiting time of the individual queues. This argument, however, assumes that the service time of the customers is perfectly scalable. Unfortunately, this assumption rarely holds in reality. In this paper we analyze the superposition of M/G/1 queues. We assume that there are scalable and unscalable components in the service time. On the basis of the criteria of average waiting time and average response time, we derive necessary and sufficient conditions for whether the superposition of queues is preferable. Examples of magnetic rotational disks are given to illustrate our findings.

This paper considers subexponential asymptotics of the tail distributions of waiting times in stationary work-conserving single-server queues with multiple Markovian arrival streams, where all arrival streams are modulated by the underlying Markov chain with finite states and service time distributions may differ for different arrival streams. Under the assumption that the equilibrium distribution of the overall (i.e., customer-average) service time distribution is subexponential, a subexponential asymptotic formula is first shown for the virtual waiting time distribution, using a closed formula recently found by the author. Further when customers are served on a FIFO basis, the actual waiting time and sojourn time distributions of customers from respective arrival streams are shown to have the same asymptotics as the virtual waiting time distribution.

The single server queue is one of the most basic queueing systems to model stochastic waiting dynamics. Most work involving the single server queue only analyzes the customer or agent behavior. However, in this work, we are inspired by COVID-19 applications and are interested in the interaction between customers and more specifically, the time that adjacent customers overlap in the queue. To this end, we derive a new recursion for this overlap time and study the steady state behavior of the overlap time via simulation and probabilistic analysis. We find that the overlap time between adjacent customers in the M/M/1 queue has a conditional distribution that is given by an exponential distribution, however, as the distance between customers grows, the probability of a non-negative overlap time decreases geometrically. We also find via simulation that the exponential distribution still holds when the distribution is non-exponential, hinting at a more general result.

Large deviation theory offers a powerful and general statistical framework to study the asymptotic dynamical properties of rare events. The application of the formalism to concrete experimental situations is, however, often restricted by finite statistics. Data might not suffice to reach the asymptotic regime or judge whether large deviation estimators converge at all. We here experimentally study the large deviation properties of the stochastic work and heat of a levitated nanoparticle subjected to nonequilibrium feedback control. This setting allows us to determine for each quantity the convergence domain of the large deviation estimators using a criterion that does not require the knowledge of the probability distribution. By extracting both the asymptotic exponential decay and the subexponential prefactors, we demonstrate that singular prefactors significantly restrict the convergence characteristics close to the singularity. Our results provide unique insight into the approach to the asymptotic large deviation limit and underscore the pivotal role of singular prefactors.

This paper studies large deviations properties of vectors of empirical means and measures generated as follows. Consider a sequence X1,X2,…,Xn of independent and identically distributed random variables partitioned into d-subgroups with sizes n1,…,nd. Further, consider a d-dimensional vector mn whose coordinates are made up of the empirical means of the subgroups. We prove the following. (1) The sequence of vector of empirical means mn satisfies large deviations principle with rate n and rate function I, when the sequence X1,X2,…,Xn is Rl valued, with l≥1. (2) Similar large deviations results hold for the corresponding sequence of vector of empirical measures Ln if Xi’s, i=1,2,…,n, take on finitely many values. (3) The rate functions for the above large deviations principles are convex combinations of the corresponding rate functions arising from the large deviations principles of the coordinates of mn and Ln. The probability distributions used in the convex combinations are given by α=(α1,…,αd)=limn→∞1/n(n1,…,nd). These results are consequently used to derive variational formula for the thermodynamic limit for the pressure of multipopulation Curie-Weiss (I. Gallo and P. Contucci (2008), and I. Gallo (2009)) and mean-field Pott’s models, via a version of Varadhan’s integral lemma for an equicontinuous family of functions. These multipopulation models serve as a paradigm for decision-making context where social interaction and other socioeconomic attributes of individuals play a crucial role.