Remaining Service Time Research Articles

The Interrelationship Between N-policy M/G/1/K and F-policy G/M/1/K Queues with Startup Time

This paper studies the interrelationship between the N-policy M/G/1/K queue and the F-policy G/M/1/K queue. The N- policy queuing problem investigates the most common issue of controlling service. The F- policy queuing problem deals with the most common issue of controlling arrivals. We provide a recursive method, using the supplementary variable technique and treating the supplementary variable as the remaining service time, to develop the solution algorithm for the N- policy M/G/1/K queue with startup time. We have demonstrated that the N-policy M/G/1/K queue with startup time has been effectively used to derive the solution algorithm to the F- policy G/M/1/K queue with startup time. The interrelationship between the N- policy and the F — policy queueing problems are illustrated analytically for 3-stage Erlang (service or interarrival) time distribution.

Stability of a retrial queueing network with different classes of customers and restricted resource pooling

We consider a retrial queueing network with different classes of customers and several servers. Each customer class is associated with a set of servers who can serve the class of customers. Customers of each class exogenously arrive according to a Poisson process. If an exogenously arriving customer finds upon his arrival any idle server who can serve the customer class, then he begins to receive a service by one of the available servers. Otherwise he joins the retrial group, and then tries his luck again after exponential time, the mean of which is determined by his customer class. Service times of each server are assumed to have general distribution. The retrial queueing network can be represented by a Markov process, with the number of customers of each class, and the customer class and the remaining service time of each busy server. Using the fluid limit technique, we find a necessary and sufficient condition for the positive Harris recurrence of the representing Markov process. This work is the first that applies the fluid limit technique to a model with retrial phenomena.

On the stationary LCFS-PR single-server queue: A characterization via stochastic intensity

We consider a stationary single-server queue with preemptive-resumelast-come, first-served (LCFS-PR) queueing discipline.The LCFS-PR single-server queue has some interesting properties andhas been studied in the literature.In this paper, we generalize the previous results such that the inputprocess to the queue is given as a general stationary marked pointprocess and derive some formulas concerning the joint distribution ofqueue length and remaining service times of respective customers inthe system at arbitrary time instances as well as at arrivalinstances.The tool for derivation is the Palm-martingale calculus; that is, theconnection between the notion of Palm probability and that ofstochastic intensity.

The Conditional Distribution of the Residual Service Time in the M n /G/1 Queue

We study a single server queueing system with general service time distribution and memoryless inter-arrival times. The arrival rate is not constant but varies with the number of customers in the system. A recursive formula for the conditional distribution of the remaining service time given the queue length is derived for an arbitrary epoch. It is also shown that this conditional distribution holds for arrival epochs. The recursion for the corresponding Laplace–Stieltjes transforms and expected values is given in a simple formula.

Optimal Control of an M/G/1/K Queueing System with Combined F Policy and Startup Time

We investigate the optimal management problem of an M/G/1/K queueing system with combined F policy and an exponential startup time. The F policy queueing problem investigates the most common issue of controlling the arrival to a queueing system. We present a recursive method, using the supplementary variable technique and treating the supplementary variable as the remaining service time, to obtain the steady state probability distribution of the number of customers in the system. The method is illustrated analytically for exponential service time distribution. A cost model is established to determine the optimal management F policy at minimum cost. We use an efficient Maple computer program to calculate the optimal value of F and some system performance measures. Sensitivity analysis is also investigated.

Stationary remaining service time conditional on queue length

In Mandelbaum and Yechiali [The conditional residual service time in the M/G/1 queue, http://www.math.tau.ac.il/ ∼ uriy/publications (No. 30a), 1979] and in Fakinos [The expected remaining service time in a single-server queue, Oper. Res. 30 (1982) 1014–1018] a simple formula is derived for the (stationary) expected remaining service time in a M/G/1 queue, conditional on the number of customers in the system. We give a short new proof of the formula using Rate Conservation Law, and generalize to handle higher moments.

Remarks on the remaining service time upon reaching a target level in the M/G/1 queue

In this note we establish connections between new and previous results on the remaining service time upon reaching a target level in the M/G/1 queue.

Product form solution for exponential G-networks with dependent service and completion of service of killed customers

Queueing networks with negative customers (G-networks), Poisson flow of positive customers, multi-server exponential nodes, and dependent service at the different nodes are studied. Every customer arriving at the network is defined by a set of random parameters: customer route, the length of customer route, customer volume and his service time at each route stage as well. A killed positive customer is removed at the last place in the queue and quits the network just after his remaining service time will be elaborated. For such G-networks, the multidimensional stationary distribution of the network state probabilities is shown to be representable in product form.

Dynamic scheduling to optimize utility functions of sojourn time moments in queueing systems

It is well known that scheduling the service of customers according to the shortest remaining processing time (SRPT) policy is optimal with respect to minimizing the mean sojourn time of customers. Recent studies have further argued that SRPT does not unfairly penalize large customers in order to benefit small customers, and therefore these studies propose the use of SRPT to improve performance in various applications. However, as Schrage and Miller point out [10], the SRPT policy can raise several difficulties for a number of important reasons. Such difficulties can arise from the inability to accurately predict service times, or the complicated nature of implementing the preemptive aspect of the SRPT policy which requires keeping track of the remaining service times of all waiting customers as well as of the customer in service. Normally, preemption also incurs additional costs. and thus one might want to avoid the preemption of customers in service whose remaining service time is not much larger than that of a new arrival.

Duration of the buffer overflow period in a batch arrival queue

The buffer overflow period is the remaining service time upon reaching a full buffer in a finite-buffer single server queueing system. Its distribution is an important characteristic of the queue, as this is the period in which arriving cells are lost. In this paper detailed studies of the buffer overflow period in a batch arrival queue are presented. Theoretical results are illustrated via numerical examples.

Nearly insensitive bounds on SMART scheduling

We define the class of SMART scheduling policies. These are policies that bias towards jobs with small remaining service times, jobs with small original sizes, or both, with the motivation of minimizing mean response time and/or mean slowdown. Examples of SMART policies include PSJF, SRPT, and hybrid policies such as RS (which biases according to the product of the remaining size and the original size of a job).For many policies in the SMART class, the mean response time and mean slowdown are not known or have complex representations involving multiple nested integrals, making evaluation difficult. In this work, we prove three main results. First, for all policies in the SMART class, we prove simple upper and lower bounds on mean response time. Second, we show that all policies in the SMART class, surprisingly, have very similar mean response times. Third, we show that the response times of SMART policies are largely insensitive to the variability of the job size distribution. In particular, we focus on the SRPT and PSJF policies and prove insensitive bounds in these cases.

DISCRETE-TIME QUEUE WITH VARIABLE SERVICE CAPACITY

This paper considers a discrete-time queueing system with variable service capacity. Using the supplementary variable method and the generating function technique, we compute the joint probability distribution of queue length and remaining service time at an arbitrary slot boundary, and also compute the distribution of the queue length at a departure time.

Formalizing SMART scheduling

It is well-known that policies which bias towards small job sizes or jobs with small remaining service times perform well with respect to mean response time and mean slowdown. This idea has been fundamental in many system implementations including the case of Web servers, where it has been shown that by giving priority to requests for small files, a Web server can significantly reduce mean response time and mean slowdown [1]. The heuristic has also been applied to other application areas; for example, scheduling in supercomputing centers.

On the Remaining Service Time upon Reaching a Given Level in M/G/1 Queues

The ability of effectively finding the distribution of the remaining service time upon reaching a target level in M/G/1 queueing systems is of great practical importance. Among other things, it is necessary for the estimation of the Quality-of-Service (QoS) provided by Asynchronous Transfer Mode (ATM) networks. The previous papers on this subject did not give a comprehensive solution to the problem. In this paper an explicit formula for this distribution is given. This formula is general as it includes any initial level of the length of the queue, any type of service distribution (heavy tails) and any traffic intensity ρ. Moreover, it is easy to use and fast in computation. To show this several numerical examples are presented. In addition, a solution of the similar problem in G/M/1 queues (which is the distribution of the remaining interarrival time) is given.

On the Queue Length Distribution for the GI/G/1/K/VM Queue

We present a transform-free distribution of the steady-state queue length for the GI/G/1/K queueing system with multiple vacations under exhaustive FIFO service discipline. The method we use is a modified supplementary variable technique and the result we obtain is expressed in terms of conditional expectations of the remaining service time, the remaining interarrival time, and the remaining vacation, conditional on the queue length at the embedded points. The case K → ∞ is also considered.

Sojourn time analysis of a two-phase queueing system with exhaustive batch-service and its vacation model

We consider an M/G/1-type, two-phase queueing system, in which the two phases in series are attended alternatively and exhaustively by a moving single-server according to a batch-service in the first phase and an individual service in the second phase. We show that the two-phase queueing system reduces to a new type of single-vacation model with non-exhaustive service. Using a double transform for the joint distribution of the queue length in each phase and the remaining service time, we derive Laplace-Stieltjes transforms for the sojourn time in each phase and the total sojourn time in the system. Furthermore, we provide the moment formula of sojourn times and numerical examples of an approximate density function of the total sojourn time.

Transient and Stationary Distributions for the GI / G / k Queue with Lebesgue-Dominated Inter-Arrival Time Distribution

In this paper, the multi-server queue with general service time distribution and Lebesgue-dominated iid inter-arival times is analyzed. This is done by introducing auxiliary variables for the remaining service times and then examining the embedded Markov chain at arrival instants. The concept of piecewise-deterministic Markov processes is applied to model the inter-arrival behaviour. It turns out that the transition probability kernel of the embedded Markov chain at arrival instants has the form of a lower Hessenberg matrix and hence admits an operator–geometric stationary distribution. Thus it is shown that matrix–analytical methods can be extended to provide a modeling tool even for the general multi-server queue.

A UNIFIED QUEUE LENGTH FORMULA FOR BMAP/G/1 QUEUE WITH GENERALIZED VACATIONS

The main purpose of this article is to demonstrate a unified queue length formula for the BMAP/G/1 queue with generalized server vacations. We also present two unified formulas; one for the distribution of the workload, and the other for the joint distribution of the queue length and the remaining service time. Some factorizations, which are previously known for M/G/1 queues with vacations and for MAP/G/1 queue with multiple vacations, are now recast in this new framework. Further, based on these formulas, we provide general recursions for the performance measures of interest.

Matrix Product-Form Solution for an LCFS-PR Single-Server Queue with Multiple Arrival Streams Governed by a Markov Chain

This paper considers a stationary single-server queue with multiple arrival streams governed by a Markov chain, where customers are served on an LCFS preemptive-resume basis. Service times of customers from each arrival stream are generally distributed and service time distributions for different arrival streams may be different. Under these assumptions, it is shown that the stationary joint distribution of queue strings representing from which arrival stream each customer in the system arrived and remaining service times of respective customers in the system has a matrix product-form solution, where matrices constituting the solution are given in terms of the infinitesimal generator of a certain Markov chain. Compared with the previous works, the result in this paper is more general in the sense that general service time distributions are allowed, and it has the advantage of computational efficiency. Note also that the result is a natural extension of the classical result for the LCFS-PR M/G/1 queue. Further, utilizing the matrix product-form solution, we derive a new expression of the vector Laplace–Stieltjes transform of the stationary distribution of unfinished work in the work-conserving single-server queue with multiple arrival streams governed by a Markov chain, which is given by the sum of matrix-geometric series.

The Remaining Service Time upon Reaching a High Level in M/G/1 Queues

The distribution of the remaining service time upon reaching some target level in an M/G/1 queue is of theoretical as well as practical interest. In general, this distribution depends on the initial level as well as on the target level, say, B. Two initial levels are of particular interest, namely, level “1” (i.e., upon arrival to an empty system) and level “B−1” (i.e., upon departure at the target level). In this paper, we consider a busy cycle and show that the remaining service time distribution, upon reaching a high level B due to an arrival, converges to a limiting distribution for B→∞. We determine this asymptotic distribution upon the “first hit” (i.e., starting with an arrival to an empty system) and upon “subsequent hits” (i.e., starting with a departure at the target) into a high target level B. The form of the limiting (asymptotic) distribution of the remaining service time depends on whether the system is stable or not. The asymptotic analysis in this paper also enables us to obtain good analytical approximations of interesting quantities associated with rare events, such as overflow probabilities.