A new method of performance sensitivity analysis for non-Markovian queueing networks

The paper develops a new method of calculating and estimating the sensitivities of a class of performance measures with respect to a parameter of the service or interarrival time distributions in queueing networks. The distribution functions may be of a general form. The study is based on perturbation analysis of queueing networks. A new concept, the realization factor of a perturbation, is introduced for the network studied. The properties of realization factors are discussed, and a set of linear differential equations specifying the realization factors are derived. The sensitivity of the steady-state performance with respect to a parameter can be expressed in a simple form using realization factors. Based on this, the sensitivity can be estimated by applying a perturbation analysis algorithm to a sample path of the system. We show that the derivative of the performance measure with respect to a parameter based on a single sample path converges with probability one to the derivative of the steady-state performance as the length of the sample path goes to infinity. The results provide a new analytical method of calculating performance sensitivities and justifies the application of perturbation analysis algorithms to non-Markovian queueing networks.     

Consistency of perturbation analysis for a queue with finite buffer space and loss policy

The subject of discrete-event dynamical systems has taken on a new direction with the advent of perturbation analysis (PA), an efficient method for estimating the gradients of a steady-state performance measure, by analyzing data obtained from a single-simulation experiment in the time domain. A crucial issue is whether PA gives strongly consistent estimates, namely, whether average time-domain-based gradients converge, over infinite horizon, to the steady-state gradients. In this paper, we investigate this issue for a queue with a finite buffer capacity and a loss policy. The performance measure in question is the average amount of lost customers, as a function of the buffer's capacity, which is assumed to be continuous in our work. It is shown that PA gives strongly consistent estimates. The analysis uses a new technique, based on busy period-dependent inequalities. This technique may have possible extensions to analyses of consistency of PA for more general queueing systems.     

Perturbation analysis and optimization of queueing networks

A new time-domain-based approach is developed in this paper for the perturbation analysis of queueing networks. We show that, by observing a single sample path realization of the network trajectory, we can derive sensitivity information of the throughput of the system with respect to various parameters. This information can then be used for the optimization of queueing networks. Numerous experiments as well as analytical results demonstrating the validity of this new approach are given and discussed.     

具有相型服务分布的排队系统的性能灵敏度分析

具有非线数服务分布的排队网络已被广泛应用于许多领域,如通讯网络和管理系统。本文借助于无穷小说矩阵摄动方法,研究了M/PH/1排队系统的稳态性能灵敏度分析问题,给出了性能灵敏度公式,并表明了稳态性能灵敏度很容易通过系统势能进行计算。同时,给出一种计算势能及性能导数的算法。这个算法可直接用于系统的控制与优化,因为它基于分析系统的一条单一样本轨道。最后提供一个数值例子来表明这个算法的应用。     

A basic formula for performance gradient estimation of semi-Markov decision processes

This paper presents a basic formula for performance gradient estimation of semi-Markov decision processes (SMDPs) under average-reward criterion. This formula directly follows from a sensitivity equation in perturbation analysis. With this formula, we develop three sample-path-based gradient estimation algorithms by using a single sample path. These algorithms naturally extend many gradient estimation algorithms for discrete-time Markov systems to continuous time semi-Markov models. In particular, they require less storage than the algorithm in the literature.     

Expansions for Joint Laplace Transform of Stationary Waiting Times in (max,+)-linear Systems with Poisson Input

We give a Taylor series expansion for the joint Laplace transform of stationary waiting times in open (max,+)-linear stochastic systems with Poisson input. Probabilistic expressions are derived for coefficients of all orders. Even though the computation of these coefficients can be hard for certain systems, it is sufficient to compute only a few coefficients to obtain good approximations (especially under the assumption of light traffic). Combining this new result with the earlier expansion formula for the mean stationary waiting times, we also provide a Taylor series expansion for the covariance of stationary waiting times in such systems.It is well known that (max,+)-linear systems can be used to represent stochastic Petri nets belonging to the class of event graphs. This class contains various instances of queueing networks like acyclic or cyclic fork-and-join queueing networks, finite or infinite capacity tandem queueing networks with various types of blocking, synchronized queueing networks and so on. It also contains some basic manufacturing models such as kanban networks, assembly systems and so forth. The applicability of this expansion technique is discussed for several systems of this type.     

考虑等待提示策略的呼叫中心流体近似方法

向顾客公布需等待的排队时间用以缓解系统拥挤是目前呼叫中心运营管理的重要手段之一。为了有效刻画等待提示策略下顾客行为变化对呼叫系统性能的影响,采用流体近似方法建立了呼叫排队系统模型。首先,通过排队分析构造等待提示影响下排队行为框架,包含带有心理行为变化特征的多种行为要素概率函数;其次,利用流体方法构建了考虑顾客重拨影响的连续排队模型,并求解了稳态条件下的系统性能指标;最后,通过与仿真模型的对比,验证了该流体近似方法的有效性与精确性。研究结果对于带有等待时间提示的呼叫中心运营具有重要指导作用。     

Robustness of perturbation analysis estimators for queueing systems with unknown distributions

Sample-path-based stochastic gradient estimators for performance measures of queueing systems rely on the assumption that a probability distribution of the random vector of interest (e.g., a service or interarrival time sequence) is given. In this paper, we address the issue of dealing with unknown probability distributions and investigate the robustness of such estimators with respect to possibly erroneous distribution choices. We show that infinitesimal perturbation analysis (IPA) can be robust in this sense and, in some cases, provides distribution-independent estimates. Comparisons with other gradient estimators are provided, including experimental results. We also show that finite perturbation analysis (FPA), though only providing gradient approximations, possesses some attractive robustness properties with respect to unknown distribution parameters. An application of FPA estimation is included for a queueing system performance optimization problem involving customers with real-time constraints.This work was supported in part by the National Science Foundation Grant ECS-88-01912 and by the Office of Naval Research Contract N00014-87-K-0304.The authors wish to thank Dr. Jack Holtzman for several useful comments and suggestions.     

非齐次特征值问题扰动分析

本文的目的是想对Ax=λx+b,x~Tx=1解的性态和扰动作一些分析,讨论解的存在范围,给出一个解扰动的估计不等式,并针对A对称情形,作较具体的分析,对大||b||和小||b||两种情形,作专门讨论.     

Stationary waiting time derivatives

We investigate the stability of waiting-time derivatives when inputs to a queueing system-service times and interarrival times-depend on a parameter. We give conditions under which the sequence of waiting-time derivatives admits a stationary distribution, and under which the derivatives converge to the stationary regime from all initial conditions. Further hypotheses ensure that the expectation of a stationary waiting-time derivative is, in fact, the derivative of the expected stationary waiting time. This validates the use of simulation-based infinitesimal perturbation analysis estimates with a variety of queueing processes.We examine waiting-time sequences satisfying recursive equations. Our basic assumption is that the input and its derivatives are stationary and ergodic. Under monotonicity conditions, the method of Loynes establishes the convergence of the derivatives. Even without such conditions, the derivatives obey a linear difference equation with random coefficients, and we exploit this fact to find stability conditions.     

A General Decomposition Algorithm for Parallel Queues with Correlated Arrivals

Queueing with correlated arrivals occurs when customers arrive at a set of queues simultaneously. The difficulty in analyzing systems with correlated arrivals is due to the fact that the individual queueing systems are stochastically dependent. Exact methods for analyzing these systems are computationally intensive and are limited to only a few special cases. In this paper, we consider a system of parallel queues with bulk service and correlated arrivals. We show how the matrix-geometric approach can be used to obtain the performance measures of the system. We also develop an algorithm for large systems that efficiently approximates the performance measures by decomposing it into individual queueing systems. Finally, we describe how the principles of our decomposition algorithm can be extended to analyze a variety of different parallel queueing systems with correlated arrivals. We then evaluate the accuracy of our algorithm through a numerical study.     

Some New Perturbation Bounds for Subunitary Polar Factors

In this paper, we present some new perturbation bounds for subunitary polar factors in a special unitarily invariant norm called a Q-norm. Some recent results in the Frobenius norm and the spectral norm are extended to the Q-norm on one hand. On the other hand we also present some relative perturbation bounds for subunitary polar factors.     

On the perturbation analysis of discrete-event dynamic systems

The paper describes a new approach to the analysis and optimization of discrete-event dynamic systems, such as queueing networks.Dedicated to G. LeitmannThe work reported in this paper was supported in part by ONR Contracts Nos. N00014-75-C-0648 and N00014-79-C-0776 and in part by NSF Grant No. NSF-ECS-82-13680.     

Error analysis for regenerative queueing estimators with special reference to gradient estimators via likelihood ratio

The performance of telecommunications systems is typically estimated (either analytically or by simulation) via queueing theoretic models. The gradient of the expected performance with respect to the various parameters (such as arrival rate or service rate) is very important as it not only measures the sensitivity to change, but is also needed for the solution of optimization problems. While the estimator for the expected performance is the sample mean of the simulation experiment, there are several possibilities for the estimator of the gradient. They include the obvious finite difference approximation, but also other recently advocated techniques, such as estimators derived from likelihood ratio transformations or from infinitesimal perturbations. A major problem in deciding upon which estimator to use and in planning the length of the simulation has been the scarcity of analytical error calculations for estimators of queueing models. It is this question that we answer in this paper for the waiting time moments (of arbitrary order) of theM / G / 1 queue by using the queueing analysis technique developed by Shalmon. We present formulas for the error variance of the estimators of expectation and of its gradient as a function of the simulation length; at arbitrary traffic intensity the formulas are recursive, while the heavy traffic approximations are explicit and of very simple form. For the gradient of the mean waiting time with respect to the arrival (or service) rate, and at 1 percent relative precision, the heavy traffic formulas show that the likelihood ratio estimator for the gradient reduces the length of the simulation required by the finite difference estimator by about one order of magnitude; further increasing the relative precision by a factor increases the reduction advantage by the same factor. At any relative precision, it exceeds the length of the simulation required for the estimation of the mean with the same relative precision by about one order of magnitude. While strictly true for theM / G / 1 queue, the formulas can also be used as guidelines in the planning of queueing simulations and of stochastic optimizations of complex queueing systems, particularly those with Poisson arivals.     

Duality for almost convex optimization problems via the perturbation approach

We deal with duality for almost convex finite dimensional optimization problems by means of the classical perturbation approach. To this aim some standard results from the convex analysis are extended to the case of almost convex sets and functions. The duality for some classes of primal-dual problems is derived as a special case of the general approach. The sufficient regularity conditions we need for guaranteeing strong duality are proved to be similar to the ones in the convex case. The research of the first and third authors was partially supported by DFG (German Research Foundation), project WA 922/1. The research of the second author was supported by the grant PN II, ID 523/2007.     

A comprehensive simplex-like algorithm for network optimization and perturbation analysis

The family of network optimization problems includes the following prototype models: assignment, critical path, max flow, shortest path, and transportation. Although it is long known that these problems can be modeled as linear programs (LP), this is generally not done. Due to the relative inefficiency and complexity of the simplex methods (primal, dual, and other variations) for network models, these problems are usually treated by one of over 100 specialized algorithms. This leads to several difficulties. The solution algorithms are not unified and each algorithm uses a different strategy to exploit the special structure of a specific problem. Furthermore, small variations in the problem, such as the introduction of side constraints, destroys the special structure and requires modifying andjor restarting the algorithm. Also, these algorithms obtain solution efficiency at the expense of managerial insight, as the final solutions from these algorithms do not have sufficient information to perform postoptimality analysis.

Another approach is to adapt the simplex to network optimization problems through network simplex. This provides unification of the various problems but maintains all the inefficiencies of simplex, as well as, most of the network inflexibility to handle changes such as side constraints. Even ordinary sensitivity analysis (OSA), long available in the tabular simplex, has been only recently transferred to network simplex.

This paper provides a single unified algorithm for all five network models. The proposed solution algorithm is a variant of the self-dual simplex with a warm start. This algorithm makes available the full power of LP perturbation analysis (PA) extended to handle optimal degeneracy. In contrast to OSA, the proposed PA provides ranges for which the current optimal strategy remains optimal, for simultaneous dependent or independent changes from the nominal values in costs, arc capacities, or suppliesJdemands. The proposed solution algorithm also facilitates incorporation of network structural changes and side constraints. It has the advantage of being computationally practical, easy for managers to understand and use, and provides useful PA information in all cases. Computer implementation issues are discussed and illustrative numerical examples are provided in the Appendix     

On the performance of damaged linear networks

The conductivity of a structured or unstructured, finite or infinite linear network consisting of nodes connected by links of varying conductance is discussed. The rate of transport of mass, heat, electricity, analogue signal, or digitized information across each link is given by the product of the link conductance and the difference in the corresponding nodal values of an appropriate driving potential. Balance equations at each node provide us with a system of linear equations whose coefficients depend on the link conductances. Link damage or disruption causes a perturbation nodal field that affects the performance of the entire network. An expression for the perturbation field relative to that established over an unperturbed network in the absence of link damage or disruption is derived based on a generalization of the Sherman–Morrison–Woodbury formula. The generalized formula provides us with the inverse of a perturbed matrix that differs from an unperturbed matrix by a sum of tensor vector products. The theoretical formulation suggests a venue for computing the effective conductance of an imperfect network and assessing critical conditions for global disruption.     

Stability analysis of parallel server systems under longest queue first

We consider the stability of parallel server systems under the longest queue first (LQF) rule. We show that when the underlying graph of a parallel server system is a tree, the standard nominal traffic condition is sufficient for the stability of that system under LQF when interarrival and service times have general distributions. Then we consider a special parallel server system, which is known as the X-model, whose underlying graph is not a tree. We provide additional “drift” conditions for the stability and transience of these queueing systems with exponential interarrival and service times. Drift conditions depend in general on the stationary distribution of an induced Markov chain that is derived from the underlying queueing system. We illustrate our results with examples and simulation experiments. We also demonstrate that the stability of the LQF depends on the tie-breaking rule used and that it can be unstable even under arbitrary low loads.     

闭排队网络中性能势与无穷小实现因子之间的关系

In this paper,the concept of the infinitesimal realization factor is extended to the parameter-dependent performance functions in closed queueing networks. Then the concepts of realization matrix (its elements are called realization factors) and performance potential are introduced,and the relations between infinitesimal realization factors and these two quantities are discussed. This provides a united framework for both IPA and non IPA approaches. Finally,another physical meaning of the service rate is given.