Variance Properties of Second Derivative Perturbation Analysis Estimators for Single-Server Queues

Fu, Michael C.,Hu, Jian-Qiang

DOI: https://doi.org/10.23919/acc.1990.4790897

1990-01-01

Abstract:We explore the variance properties of three second derivative perturbation analysis estimators for single-server queues.

What problem does this paper attempt to address?

Comparison of Gradient Estimation Techniques for Queues with Non-Identical Servers.

MC FU,JQ HU,R NAGI

DOI: https://doi.org/10.1016/0305-0548(94)00065-g

IF: 5.159

1995-01-01

Computers & Operations Research

Abstract:Gradient estimation techniques for stochastic discrete-event simulation have been a major topic of research over the past decade. In this paper, we apply two of the techniques-perturbation analysis and the likelihood ratio method-to a single-queue system with non-identical multiple servers. We derive estimates For derivatives of mean steady-state system time with respect to parameters of the underlying timing distributions. In terms of perturbation analysis, we consider both an infinitesimal perturbation analysis estimator, which is biased for this problem, and two smoothed perturbation analysis estimators, one unbiased but not very practical and one approximate but easily implementable. For two servers, we provide an analytical proof of unbiasedness in steady state for the Markovian case. For the likelihood ratio method, we apply the regenerative likelihood ratio estimator. We provide simulation results for both Markovian and non-Markovian examples, and compare the performance of the various estimators. We conclude that no one method performs universally well, and provide recommendations as to when one is likely to be preferred to the others.
Second derivative sample path estimators for the GI/G/m queue

Michael C. Fu,Jian-Qiang Hu

DOI: https://doi.org/10.1287/mnsc.39.3.359

IF: 5.4

1993-01-01

Management Science

Abstract:Applying the technique of smoothed perturbation analysis (SPA) to the GI / G / m queue with first-come, first-served (FCFS) queue discipline, we derive sample path estimators for the second derivative of mean steady-state system time with respect to a parameter of the service time distribution. Such estimators provide a possible means for speeding up the convergence of gradient-based stochastic optimization algorithms. The derivation of the estimators sheds some new light on the complications encountered in applying the technique of SPA. The most general cases require the simulation of additional sample subpaths; however, an approximation procedure is also introduced which eliminates the need for additional simulation. Simulation results indicate that the approximation procedure is reasonably accurate. When the service times are exponential or deterministic, the estimator simplifies and the approximation procedure becomes exact. For the M / M / 2 queue, the estimator is proved to be strongly consistent.
The Variance of Partial Sums of Strong Near-Epoch Dependent Variables

Jin Qiu,Zhengyan Lin

DOI: https://doi.org/10.1016/j.spl.2006.04.031

IF: 0.718

2006-01-01

Statistics & Probability Letters

Abstract:In this paper, we discuss the lower bound for the variance of partial sums of strong near-epoch dependent variables under quite weak conditions. By suitably applying this result, we derive the limit behavior of the variance of the partial sums, which is especially useful in studying the large sample properties for econometric time series models with strong near-epoch dependent innovations.
On Perturbation Propagation for Smoothed Perturbation Analysis

Michael C. Fu,Jian-Qiang Hu

DOI: https://doi.org/10.1016/s1474-6670(17)58438-4

1996-01-01

IFAC Proceedings Volumes

Abstract:A critical component in the implementation of smoothed perturbation analysis (SPA) estimators is the propagation of conditional perturbations. Left- and right-hand second derivative estimators for a single-server queue are used to illustrate that one often has to be careful when it comes to perturbation propagation. In this particular example, the propagation can be carried out for the right-hand estimator using the original sample path, but not for the left-hand estimator. However, for this example, the difficulty can be overcome by simplifying the original left-hand derivative estimator so that no propagation is needed.
Nonparametric Estimation of Quantiles for A Class of Stationary Processes

Chu Huang,HanChao Wang,ZhengYan Lin

DOI: https://doi.org/10.1007/s11425-015-5024-2

2015-01-01

Science China Mathematics

Abstract:We study smoothed quantile estimator for a class of stationary processes. We obtain the convergency rates and the Bahadur representation, as well as the asymptotic normality for this estimator by the method of m-dependent approximation. Our results can be used in the study of the estimation of value-at-risk (VaR) and applied to many time series which have important applications in econometrics.
On the Variance of Single-Run Unbiased Stochastic Derivative Estimators.

Zhenyu Cui,Michael C. Fu,Jian-Qiang Hu,Yanchu Liu,Yijie Peng,Lingjiong Zhu

DOI: https://doi.org/10.1287/ijoc.2019.0897

IF: 3.288

2019-01-01

INFORMS Journal on Computing

Abstract:We analyze the variance of single-run unbiased stochastic derivative estimators. The distribution of a specific conditional expectation characterizes an intrinsic distributional property of the derivative estimators in a given class, which, in turn, separates two of the most popular single-run unbiased derivative estimators, infinitesimal perturbation analysis and the likelihood ratio method, into disjoint classes. In addition, a necessary and sufficient condition for the estimators to achieve the lowest variance in a certain class is provided, as well as insights into finding an estimator with lower variance. We offer a sufficient condition to substantiate the rule of thumb that the infinitesimal perturbation analysis estimator has a smaller variance than does the likelihood ratio method estimator and to provide a counterexample when the sufficient condition is not satisfied.
Smoothed Perturbation Analysis for Queues with Finite Buffers

M. C. Fu,J. Q. Hu

DOI: https://doi.org/10.1007/bf01153526

IF: 1.402

1993-01-01

Queueing Systems

Abstract:Applying the technique of smoothed perturbation analysis (SPA) to theGI/G/1/K queue, we derive gradient estimators for two performance measures: the mean steady-state system time of a served customer and the probability that an arriving customer is rejected. Unbiasedness of the estimators follows from results of a previous general framework on SPA estimators. However, in that framework, the estimators often require the simulation of numerous additional sample subpaths, possibly making the technique practically infeasible in applications. We exploit some of the special structure of theGI/G/1/K queue to come up with an estimator which requires at most the simulation of a single additional sample subpath. By establishing certain regenerative properties, we provide a strong consistency proof for the estimator.
A Bound on the Variance of the Waiting Time in a Queueing System

Patrick Eschenfeldt,Ben Gross,Nicholas Pippenger

DOI: https://doi.org/10.48550/arXiv.1106.0074

2011-06-01

Abstract:Kingman has shown, under very weak conditions on the interarrival- and sevice-time distributions, that First-Come-First-Served minimizes the variance of the waiting time among possible service disciplines. We show, under the same conditions, that Last-Come-First-Served maximizes the variance of the waiting time, thereby giving an upper bound on the variance among all disciplines.

Probability
Bias Properties Of Infinitesimal Perturbation Analysis For Multiserver Queues

Mc Fu,Jq Hu

DOI: https://doi.org/10.1109/WSC.1990.129545

1990-01-01

Abstract:Infinitesimal perturbation analysis (IPA) gives biased estimates for multi-server queues when the service time distributions of the servers are not equal. The authors analytically calculate the bias for IPA gradient estimators of mean steady-state system time (with respect to parameters in the inter-arrival or service time distributions) in a two-server Markovian queue where reversibility holds. Because it is believed that the applicability of IPA is in general distribution-independent, it is conjectured that the order of biases with respect to nonidenticality may be applicable to more general distributions. This would mean that IPA is relatively insensitive to small deviations from the equal server case. >
Sample path properties of the G/D/m queue

michael c fu,jianqiang hu

DOI: https://doi.org/10.1016/0377-2217(93)90339-O

IF: 6.4

1993-01-01

European Journal of Operational Research

Abstract:Infinitesimal perturbation analysis is a technique for estimating gradients of performance measures in stochastic discrete-event systems - which are often used to model computer/communications networks and manufacturing systems - from a single simulation of the system. In Fu and Hu (1991), the application of infinitesimal perturbation analysis to multi-server queues was developed, and the consistency of the estimates was investigated, with an analytical proof provided for the exponential case (the M/M/m queue) and a heuristic limit argument presented for the general case. Here, we use sample path analysis to give a consistency proof for the case of deterministic service times (the G/D/m queue). The key result needed for the proof is convexity of the sample performance.
On Comparison of Steady-State Infinitesimal Perturbation Analysis and Likelihood Ratio Derivative Estimates.

Jian-Qiang Hu,Teng Lian

DOI: https://doi.org/10.1007/s10626-022-00370-8

2023-01-01

Discrete Event Dynamic Systems

Abstract:In this paper, we compare the infinitesimal perturbation analysis (IPA) and likelihood ratio (LR) derivative estimators for the steady-state system time of the M/M/1 queue. We derive the explicit formulas for the variances of both estimators. We then show IPA has smaller variance, which is often considered as a rule-of-thumb in the simulation literature. This is the first theoretical result that proves IPA has smaller variance than LR for steady-state performance measures.
Bias Properties Of Infinitesimal Perturbation Analysis For Systems With Parallel Servers

Mc Fu,Jq Hu,R Nagi

DOI: https://doi.org/10.1016/0305-0548(92)90070-L

IF: 5.159

1992-01-01

Computers & Operations Research

Abstract:Infinitesimal perturbation analysis (IPA) is a technique for estimating derivatives of performance measures from a single simulation of a stochastic discrete-event system, which might, for example, be modeling a computer/communication network or a manufacturing system. Such derivative estimates are useful in sensitivity analysis of the system and in optimizing-or at least improving-the performance of the system through gradient algorithms. However, it is well-known that IPA gives biased estimates for systems with multiple servers (channels or machines) in parallel when the server characteristics are not identical. In this paper, we investigate the seriousness of the bias both theoretically and experimentally. From our study, we believe the IPA estimator is relatively insensitive to small differences in the service time distributions of the servers. Specifically, we conjecture that for nearly-identical servers, the bias is proportional to the square of the difference between the means of the service time distributions. To support this conjecture, we present a tractable analytical example and investigate, via simulation experiments, more general systems.
Second-order bounds for the M/M/$s$ queue with random arrival rate

Wouter J.E.C. van Eekelen,Grani A. Hanasusanto,John J. Hasenbein,Johan S.H. van Leeuwaarden

2023-10-16

Abstract:Consider an M/M/$s$ queue with the additional feature that the arrival rate is a random variable of which only the mean, variance, and range are known. Using semi-infinite linear programming and duality theory for moment problems, we establish for this setting tight bounds for the expected waiting time. These bounds correspond to an arrival rate that takes only two values. The proofs crucially depend on the fact that the expected waiting time, as function of the arrival rate, has a convex derivative. We apply the novel tight bounds to a rational queueing model, where arriving individuals decide to join or balk based on expected utility and only have partial knowledge about the market size.

Probability,Optimization and Control
Smoothed Perturbation Analysis Derivative Estimation For Markov-Chains

Mc Fu,Jq Hu

DOI: https://doi.org/10.1016/0167-6377(94)90084-1

IF: 1.151

1994-01-01

Operations Research Letters

Abstract:Using the technique of smoothed perturbation analysis, we consider steady-state performance measures for continuous-time Markov chains and derive a derivative estimator that can be estimated from a single sample path of the chain under consideration. The estimator is applicable to multi-class queueing networks, for which previous infinitesimal perturbation analysis estimators failed. A simple multi-class queueing network is used to illustrate the application of the estimator.
Theory and Methodology Higher order approximations for the single server queue with splitting, merging and feedback

Muckai K. Girish,Jian-Qiang Hu

DOI: https://doi.org/10.1016/S0377-2217(99)00174-5

IF: 6.4

2000-01-01

European Journal of Operational Research

Abstract:The performance evaluation of many modern communication and manufacturing systems has been made possible by modeling them as queueing networks. Due to the limitations of exact algorithms to analyze general queueing networks, researchers have been focusing on developing approximations. Most of the approximations are lower order approxi- mations in the sense that only a few parameters of the interarrival and service time distributions are used. This could lead to large errors in many conditions. In this paper, higher order approximations are developed for the single server queue with splitting, merging and feedback phenomena, which can then be integrated along with parametric decom- position to develop higher order approximations for queueing networks. Not only are the higher order approximations more accurate, but they can also be used to estimate the higher moments of the performance measures. " 2000 Elsevier Science B.V. All rights reserved.
Consistency of infinitesimal perturbation analysis for the GI/G/m queue

michael c fu,jianqiang hu

DOI: https://doi.org/10.1016/0377-2217(91)90328-S

IF: 6.4

1991-01-01

European Journal of Operational Research

Abstract:We extend the infinitesimal perturbation analysis algorithm developed for the GI/G/1 queue to the GI/G/m queue. The primary difference is the need for additional accumulators for each server. For the M/M/2 queue, a consistency proof is given which is extendable to the M/M/m queue. For the more general GI/G/m queue, a heuristic argument is given as to why the algorithm works. In this argument, it is seen that the need for the servers to be identical is crucial for the correctness of the algorithm. Numerical results are provided.
Extensions and Generalizations of Smoothed Perturbation Analysis in a Generalized Semi-Markov Process Framework

MC FU,JQ HU

DOI: https://doi.org/10.1109/9.256367

IF: 6.549

1992-01-01

IEEE Transactions on Automatic Control

Abstract:Under a very general framework, both in terms of finite-time performance measures and system structure, the authors derive smoothed perturbation analysis (SPA) estimators and prove their unbiasedness. The commuting condition, which has been key in previous work, is not required a priori, and thus the framework includes such systems as the GI/G/1/K queue and multiclass queueing networks, which do not satisfy the commuting condition. The generality achieved is traded off against the fact that the estimator is not always easily implementable on a single sample path. The use of the commuting condition in a local sense is proposed to help simplify the estimators derived: queueing and multiclass queueing networks are used as illustrative examples. For a simple multiclass closed queueing network, some simulation results are provided. When the commuting condition is satisfied globally, the framework allows the recovery of previous results on IPA and SPA estimators as corollaries of the main theorems. >
Perturbation Theory for the Asymptotic Decay Rates in the Queues with Markovian Arrival Process and/or Markovian Service Process – Correction

Attahiru Sule Alfa,Jungong Xue,Qiang Ye

DOI: https://doi.org/10.1023/A:1011032718715

2001-01-01

Abstract:Three kinds of queues with Markovian arrival process and/or Markovian service process, are considered in this paper. In great generality, their basic steady-state distributions have asymptotically exponential tails. We investigate the sensitivity of these asymptotic decay rates to the small entrywise perturbations in the parameter matrices of the Markovian arrival process.
Many Server Queueing Models with Heterogeneous Servers and Parameter Uncertainty with Customer Contact Centre Applications

Wenyi Qin

DOI: https://doi.org/10.48550/arXiv.1812.00341

2018-12-02

Abstract:In this thesis, we study the queueing systems with heterogeneous servers and service rate uncertainty under the Halfin-Whitt heavy traffic regime. First, we analyse many server queues with abandonments when service rates are i.i.d. random variables. We derive a diffusion approximation using a novel method. The diffusion has a random drift, and hence depending on the realisations of service rates, the system can be in Quality Driven (QD), Efficiency Driven (ED) or Quality-Efficiency-Driven (QED) regime. When the system is under QD or QED regime, the abandonments are negligible in the fluid limit, but when it is under ED regime, the probability of abandonment will converge to a non-zero value. We then analyse the optimal staffing levels to balance holding costs with staffing costs combining these three regimes. We also analyse how the variance of service rates influence abandonment rate. Next, we focus on the state space collapse (SSC) phenomenon. We prove that under some assumptions, the system process will collapse to a lower dimensional process without losing essential information. We first formulate a general method to prove SSC results inside pools for heavy traffic systems using the hydrodynamic limit idea. Then we work on the SSC in multi-class queueing networks under the Halfin-Whitt heavy traffic when service rates are i.i.d. random variables within pools. For such systems, exact analysis provides limited insight on the general properties. Alternatively, asymptotic analysis by diffusion approximation proves to be effective. Further, limit theorems, which state the diffusively scaled system process weakly converges to a diffusion process, are usually the central part in such asymptotic analysis. The SSC result is key to proving such a limit. We conclude by giving examples on how SSC is applied to the analysis of systems.

Probability
Consistency of infinitesimal perturbation analysis estimators with rates

Jian-Qiang HU

DOI: https://doi.org/10.1007/BF02412255

IF: 1.402

1991-01-01

Queueing Systems

Abstract:We study a class of infinitesimal perturbation analysis (IPA) algorithms for queueing systems with load-dependent service and/or arrival rates. Such IPA algorithms were originally motivated by applications to large queueing systems in conjunction with aggregation algorithms. We prove strong consistency of these estimators through a type of birth and death queue.

Variance Properties of Second Derivative Perturbation Analysis Estimators for Single-Server Queues

Comparison of Gradient Estimation Techniques for Queues with Non-Identical Servers.

Second derivative sample path estimators for the GI/G/m queue

The Variance of Partial Sums of Strong Near-Epoch Dependent Variables

On Perturbation Propagation for Smoothed Perturbation Analysis

Nonparametric Estimation of Quantiles for A Class of Stationary Processes

On the Variance of Single-Run Unbiased Stochastic Derivative Estimators.

Smoothed Perturbation Analysis for Queues with Finite Buffers

A Bound on the Variance of the Waiting Time in a Queueing System

Bias Properties Of Infinitesimal Perturbation Analysis For Multiserver Queues

Sample path properties of the G/D/m queue

On Comparison of Steady-State Infinitesimal Perturbation Analysis and Likelihood Ratio Derivative Estimates.

Bias Properties Of Infinitesimal Perturbation Analysis For Systems With Parallel Servers

Second-order bounds for the M/M/$s$ queue with random arrival rate

Smoothed Perturbation Analysis Derivative Estimation For Markov-Chains

Theory and Methodology Higher order approximations for the single server queue with splitting, merging and feedback

Consistency of infinitesimal perturbation analysis for the GI/G/m queue

Extensions and Generalizations of Smoothed Perturbation Analysis in a Generalized Semi-Markov Process Framework

Perturbation Theory for the Asymptotic Decay Rates in the Queues with Markovian Arrival Process and/or Markovian Service Process – Correction

Many Server Queueing Models with Heterogeneous Servers and Parameter Uncertainty with Customer Contact Centre Applications

Consistency of infinitesimal perturbation analysis estimators with rates