Asymptotic analysis for closed multiclass queueing networks in critical usage

We consider a class of closed multiclass queueing networks containing First-Come-First-Serve (FCFS) and Infinite Server (IS) stations. These networks have a productform solution for their equilibrium probabilities. We study these networks in an asymptotic regime for which the number of customers and the service rates at the FCFS stations go to infinity with the same order. We assume that the regime is in critical usage, whereby the utilizations of the FCFS servers slowly approach one. The asymptotic distribution of the normalized queue lengths is shown to be in many cases a truncated multivariate normal distribution. Traffic conditions for which the normalized queue lengths arealmost asymptotically independent are determined. Asymptotic expansions of utilizations and expected queue lengths are presented. We show through an example how to obtain asymptotic expansions of performance measures when the networks are in mixed usage and how to apply the results to networks with finite data.Supported partially by NSF grant NCR93-04601.     

Instability of LAS multiclass queueing networks

We provide an example of a strictly subcritical multiclass queueing network which is unstable under the least attained service (LAS) service protocol. It is a reentrant line with two servers and six customer classes. The customer interarrival times in our system are bounded below and have finite exponential moments, while the corresponding service times are deterministic. As a special case, we obtain a deterministic, strictly subcritical unstable LAS network.     

Value iteration and optimization of multiclass queueing networks

This paper considers in parallel the scheduling problem for multiclass queueing networks, and optimization of Markov decision processes. It is shown that the value iteration algorithm may perform poorly when the algorithm is not initialized properly. The most typical case where the initial value function is taken to be zero may be a particularly bad choice. In contrast, if the value iteration algorithm is initialized with a stochastic Lyapunov function, then the following hold: (i) a stochastic Lyapunov function exists for each intermediate policy, and hence each policy is regular (a strong stability condition), (ii) intermediate costs converge to the optimal cost, and (iii) any limiting policy is average cost optimal. It is argued that a natural choice for the initial value function is the value function for the associated deterministic control problem based upon a fluid model, or the approximate solution to Poisson’s equation obtained from the LP of Kumar and Meyn. Numerical studies show that either choice may lead to fast convergence to an optimal policy. This revised version was published online in June 2006 with corrections to the Cover Date.     

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.     

Simple bounds for closed queueing networks

Consider a closed Jackson type network in which each queue has a single exponential server. Assume that N customers are moving among k queues. We establish simple closed form bounds on the network throughput (both lower and upper), which are sharper than those that are currently available. Numerical evaluation indicates that the improvements are significant.     

State space collapse with application to heavy traffic limits for multiclass queueing networks

Heavy traffic limits for multiclass queueing networks are a topic of continuing interest. Presently, the class of networks for which these limits have been rigorously derived is restricted. An important ingredient in such work is the demonstration of state space collapse. Here, we demonstrate state space collapse for two families of networks, first-in first-out (FIFO) queueing networks of Kelly type and head-of-the-line proportional processor sharing (HLPPS) queueing networks. We then apply our techniques to more general networks. To demonstrate state space collapse for FIFO networks of Kelly type and HLPPS networks, we employ law of large number estimates to show a form of compactness for appropriately scaled solutions. The limits of these solutions are next shown to satisfy fluid model equations corresponding to the above queueing networks. Results from Bramson [4,5] on the asymptotic behavior of these limits then imply state space collapse. The desired heavy traffic limits for FIFO networks of Kelly type and HLPPS networks follow from this and the general criteria set forth in the companion paper Williams [41]. State space collapse and the ensuing heavy traffic limits also hold for more general queueing networks, provided the solutions of their fluid model equations converge. Partial results are given for such networks, which include the static priority disciplines. This revised version was published online in June 2006 with corrections to the Cover Date.     

Deadlock free buffer allocation in closed queueing networks

Blocking queueing networks are of much interest in performance analysis due to their realistic modeling capability. One important feature of such networks is that they may have deadlocks which can occur if the node capacities are not sufficiently large. A necessary and sufficient condition for the node capacities is presented such that the network is deadlock free. An algorithm is given for buffer allocation in blocking queueing networks such that no deadlocks will occur assuming that the network has the special structure called cacti-graph. Additional algorithm which takes linear time in the number of nodes, is presented to find cycles in cacti networks.Akyildiz's work was supported in part by School of Information and Computer Science, ICS, of Georgia Tech and by the Air Force Office of the Scientific Research (AFOSR) under Grant AFOSR-88-0028.     

Near-optimal analysis of homogeneous central-server queueing networks

The problem of achieving near-optimality in homogeneous central-server queueing networks is investigated by means of a composite approach based on approximate operational analysis and goal programming procedures. A near-optimal solution is shown to exist: this includes the expected overall waiting time, the overall throughput rate, as well as the distribution of queue length values. The need to maintain a balanced network flow and the desire to minimize the overall waiting time are expressed as complimentary objectives. Numerical results, based on past measurements from a multi-server computing facility, indicate that the performance gains obtained by the application of the present methodology are quite significant throughout the network's feasible population scale. © 1997 by John Wiley & Sons, Ltd.     

Mean-value analysis for a class of Petri nets and batch-movement queueing networks with product-form equilibrium distributions

A number of recent papers have shown that there are classes of queueing networks, with batches of customers served and routed through the network, which have generalized product form equilibrium distributions. This extends to some Petri nets. In this paper, we indicate how a class of these is amenable to a mean-value analysis similar to that used for single-movement networks. To bring out the simplicity of the underlying ideas, we do this by working a simple example rather than presenting the development in its generality.     

Dynamic scheduling in multiclass queueing networks: Stability under discrete-review policies

This paper describes a family of discrete-review policies for scheduling open multiclass queueing networks. Each of the policies in the family is derived from what we call a dynamic reward function: such a function associates with each queue length vector q and each job class k a positive value r _k (q), which is treated as a reward rate for time devoted to processing class k jobs. Assuming that each station has a traffic intensity parameter less than one, all policies in the family considered are shown to be stable. In such a policy, system status is reviewed at discrete points in time, and at each such point the controller formulates a processing plan for the next review period, based on the queue length vector observed. Stability is proved by combining elementary large deviations theory with an analysis of an associated fluid control problem. These results are extended to systems with class dependent setup times as well as systems with alternate routing and admission control capabilities. This revised version was published online in June 2006 with corrections to the Cover Date.     

Branching/queueing networks: Their introduction and near-decomposability asymptotics

A new class of models, which combines closed queueing networks with branching processes, is introduced. The motivation comes from MIMD computers and other service systems in which the arrival of new work is always triggered by the completion of former work, and the amount of arriving work is variable. In the variant of branching/queueing networks studied here, a customer branches into a random and state-independent number of offspring upon completing its service. The process regenerates whenever the population becomes extinct. Implications for less rudimentary variants are discussed. The ergodicity of the network and several other aspects are related to the expected total number of progeny of an associated multitype Galton-Watson process. We give a formula for that expected number of progeny. The objects of main interest are the stationary state distribution and the throughputs. Closed-form solutions are available for the multi-server single-node model, and for homogeneous networks of infinite-servers. Generally, branching/queueing networks do not seem to have a product-form state distribution. We propose a conditional product-form approximation, and show that it is approached as a limit by branching/queueing networks with a slowly varying population size. The proof demonstrates an application of the nearly complete decomposability paradigm to an infinite state space. This revised version was published online in June 2006 with corrections to the Cover Date.     

Symmetry property of the throughput in closed tandem queueing networks with finite buffers

In this paper we consider closed tandem queueing networks with finite buffers and blocking before service. With this type of blocking, a server is allowed to start processing a job only if there is an empty space in the next buffer. It was recently conjectured that the throughput of such networks is symmetrical with respect to the population of the network. That is, the throughput of the network with population N is the same as that with population C − N, where C is the total number of buffer spaces in the network. The main purpose of this paper is to prove this result in the case where the service time distributions are of phase type (PH-distribution). The proof is based on the comparison of the sample paths of the network with populations N and C − N. Finally, we also show that this symmetry property is related to a reversibility property of this class of networks.     

A general formulation for mean-value analysis in product-form batch-movement queueing networks

A number of recent papers have shown that many classes of queueing networks with batches of customers served and routed through the network have equilibrium distributions which factorise into product forms over the nodes of the network. In this paper we demonstrate how such networks are amenable to a mean-value analysis which generalises that used for single-movement networks.Since product-form stochastic Petri nets (SPNs) can be viewed as batch-movement queueing networks, our algorithm is also applicable to their analysis.     

Estimating the sojourn time sensitivity in queueing networks using perturbation analysis

The sample path perturbation analysis technique developed earlier for the analysis of throughput sensitivities (Refs. 1–3) is extended to the performance measures involving mean sojourn times of customers. The major features of the sojourn time sensitivity problem are twofold. Firstly, it is a performance associated with servers, and not with customers. Secondly, the average sojourn time in any finite observation period can be a discontinuous function of mean service times when blocking is involved in a system. This discontinuity causes errors which must be accounted for in the estimation of sensitivities. Numerical experiments and analysis validate this method of computation of the sensitivities.This work was supported by the US Office of Naval Research, Contracts N00014-84-K-0465 and N00014-79-C-0776, and by the National Science Foundation, Grant No. ENG-78-15231.     

A large closed queueing network with autonomous service and bottleneck

This paper studies the queue-length process in a closed Jackson-type queueing network with the large number N of homogeneous customers by methods of the theory of martingales and by the up- and down-crossing method. The network considered here consists of a central node (hub), being an infinite-server queueing system with exponentially distributed service times, and k single-server satellite stations (nodes) with generally distributed service times with rates depending on the value N. The service mechanism of these k satellite stations is autonomous, i.e., every satellite server j serves the customers only at random instants that form a strictly stationary and ergodic sequence of random variables. Assuming that the first k-1 satellite stations operate in light usage regime the paper considers the cases where the kth satellite station is a bottleneck node. The approach of the paper is based both on development of the method from the paper by Kogan and Liptser [16], where a Markovian version of this model has been studied, and on development of the up- and down-crossing method. This revised version was published online in June 2006 with corrections to the Cover Date.     

A class of hierarchical queueing networks and their analysis

Queueing networks are an adequate model type for the analysis of complex system behavior. Most of the more realistic models are rather complex and do not fall into the easy solvable class of product form networks. Those models have to be analyzed by numerical solution of the underlying Markov chain and/or approximation techniques including simulation. In this paper a class of hierarchically structured queueing networks is considered and it is shown that the hierarchical model structure is directly reflected in the state space and the generator matrix of the underlying Markov chain. Iterative solution techniques for stationary and transient analysis can be modified to make use of the model structure and allow an efficient numerical analysis of large, up to now not solvable queueing networks.     

Smoothed perturbation analysis algorithms for estimating the derivatives of occupancy-related functions in serial queueing networks

We present unbiased Smoothed Perturbation Analysis (SPA) estimators for the derivatives of occupancy-related performance functions in serial networks ofG/G/1 queues with respect to parameters of the distributions of service times at the queues. The sample functions for these performance measures are piecewise constant, and established Infinitesimal Perturbation Analysis (IPA) methods typically fail to provide unbiased estimators in this case. The performance measures considered in this paper are: the average network occupancy as seen by an arrival, the average occupancy of a specific queue as seen by an arrival to it, the probability that a customer is blocked at a specific queue, and the probability that a customer leaves a queue idle. The SPA estimators derived are quite simple and flexible, and they lend themselves to straightforward analysis. Unlike most of the established SPA algorithms, ours are not based on the comparison of hazard rates, and the proofs of their unbiasedness do not require the boundedness of such hazard rates.Supported in part by the National Science Foundation under Grant ECS-8801912, by the Office of Naval Research under Contract N00014-87-K-0304, and by NASA under Contract NAG 2-595.     

Review of a fast simulation method for the analysis of queueing networks

This paper presents a review of and refinements to a class of discrete-event models for the analysis of unreliable queueing systems. In contrast to conventional piece-by-piece simulators, these models observe a number of rare events that affect the inflow and outflow rates at each queue. Between events, the evolution of the system is approximated by a linear function. Several experiments confirm the accuracy of this approximation and its computational efficiency over conventional simulation. © 1997 by John Wiley & Sons, Ltd.     

Piecewise linear test functions for stability and instability of queueing networks

We develop the use of piecewise linear test functions for the analysis of stability of multiclass queueing networks and their associated fluid limit models. It is found that if an associated LP admits a positive solution, then a Lyapunov function exists. This implies that the fluid limit model is stable and hence that the network model is positive Harris recurrent with a finite polynomial moment. Also, it is found that if a particular LP admits a solution, then the network model is transient. This revised version was published online in June 2006 with corrections to the Cover Date.     

Towards better multi-class parametric-decomposition approximations for open queueing networks

Methods are developed for approximately characterizing the departure process of each customer class from a multi-class single-server queue with unlimited waiting space and the first-in-first-out service discipline. The model is (GT _i /GI _i )/1 with a non-Poisson renewal arrival process and a non-exponential service-time distribution for each class. The methods provide a basis for improving parametric-decomposition approximations for analyzing non-Markov open queueing networks with multiple classes. For example, parametric-decomposition approximations are used in the Queueing Network Analyzer (QNA). The specific approximations here extend ones developed by Bitran and Tirupati [5]. For example, the effect of class-dependent service times is considered here. With all procedures proposed here, the approximate variability parameter of the departure process of each class is a linear function of the variability parameters of the arrival processes of all the classes served at that queue, thus ensuring that the final arrival variability parameters in a general open network can be calculated by solving a system of linear equations.