Showing papers on "Server published in 1979"
TL;DR: It is shown that during periods when most servers are down, large build-ups may occur which affect the queue adversely for a long time afterwards, and it is found that such build-up are aggravated by reducing the number of repair crews, but may be attenuated by reduced the arrival rates during periodsWhen the service has deteriorated.
Abstract: We study a queue with N servers, who may break down and require repair at a facility, which has c repair crews. Under exponential assumptions, this model has an algorithmically tractable solution. It is then in fact a particular case of the M/M/N queue in a Markovian environment. The purpose of this paper is twofold. As a novel methodological contribution, the stationary distributions of various waiting times are discussed. Although fairly involved, these distributions may be computed by classical numerical methods. The second and primary purpose is to demonstrate the utility of interactive computation in answering questions on the behavior, design and control of certain service systems. By numerical examples, we shall show that during periods when most servers are down, large build-ups may occur which affect the queue adversely for a long time afterwards. We also find that such build-ups are aggravated by reducing the number of repair crews, but may be attenuated by reducing the arrival rates during peri...
154 citations
TL;DR: A message switched network is used to solve a problem that has always been solved in a highly sequential fashion and the specific problem that is studied is discrete-event simulation though key concepts can be extended to other areas of message-switched problem-solving.
65 citations
[...]
TL;DR: In the formulation of I, even the system with a single server is non-trivial for the authors are concerned not just with the queue length at the first server but with the joint probability distribution of D0(t) and D1(t).
Abstract: In the formulation of I, even the system with a single server is non-trivial for we are concerned not just with the queue length at the first server but with the joint probability distribution of D0(t) and D1(t), or equivalently, the joint distribution of D0(t) or D1(t) and the queue length D0(t) − D1(t).
42 citations
TL;DR: Three conclusions of interest to the practicing manager are the primary result of the simulation: increasing server transfer time increases mean flow time, flow time variance, and the percent of time the server spends transferring.
Abstract: Simulation is used to evaluate a number of server assignment prodecures within a system of parallel queues. There are less servers than parallel single-channel service facilities and server transfer between facilities is allowed. Server transfer involves a deterministic time delay. Job interarrival times and service times follow the negative exponential distribution. Evaluated assignment procedures include, (1) rules incorporating parametrically weighted transfer times, and (2) rules permitting transfer only when a fixed minimum amount of time has elapsed. These rules have the effect of delaying server assignment. Three conclusions of interest to the practicing manager are the primary result of the simulation: (1) Increasing server transfer time increases mean flow time, flow time variance, and the percent of time the server spends transferring. (2) Using rules to delay server assignment reduces these criteria when compared with rules which do not delay assignment. A manager should consider delaying assig...
37 citations
TL;DR: In this article, the authors compare two queueing systems with identical general arrival streams, but different numbers of servers, different waiting room capacities, and stochastically ordered service time distributions.
Abstract: We compare two queueing systems with identical general arrival streams, but different numbers of servers, different waiting room capacities, and stochastically ordered service time distributions. Under appropriate conditions, it is possible to construct two new systems on the same probability space so that the new systems are probabilistically equivalent to the original systems and each sample path of the stochastic process representing system size in one system lies entirely below the corresponding sample path in the other system. This construction implies stochastic order for these processes and many associated quantities of interest, such as a busy period, the number of customers lost in any interval, and the virtual waiting time.
37 citations
TL;DR: In this paper, the authors describe a time-dependent queueing model of the helitack system and use numerical methods to estimate some of its operating characteristics, and then use a dynamic programming model to specify an optimal allocation of the available helicopters to helitacking bases.
Abstract: Helicopters are used extensively to transport initial-attack crews to forest fires in the province of Ontario. Each day fire managers must decide how to allocate the available helicopters to initial-at.ta ck bases. The helitack transport system at each base can be viewed as a multi-channel queue with customers (fires) and servers (helicopters). The authors describe a time-dependent queueing model of the helitack system and use numerical methods to estimate some of its operating characteristics. A dynamic programming model is then used to specify an optimal allocation of the available helicopters to helitack bases.
30 citations
PARC1
TL;DR: A file servers commonly have a user interface in which the client identifies himself, authenticates himself with a password, and then has access to certain files as indicated in a directory system including access controls of various sorts.
Abstract: A fi le server is a computer whose purpose is to maintain a store of files accessible via some kind of network. It may be solely dedicated to this purpose, or it may do other things too. Such file servers commonly have a user interface in which the client identifies himself, authenticates himself with a password, and then has access to certain files as indicated in a directory system including access controls of various sorts. There is very commonly a requirement that all persons wishing to have access of any kind to the file server must be registered users of it; this requirement cannot easily be met in a large network connecting file servers and other machines controlled by different adminstrative authorities.
26 citations
TL;DR: It is shown that a particular weighted inverse combination of Erlang B and C formulas, herein called the hybrid, is useful for grade of service estimates under several conditions.
Abstract: This paper attempts to broaden the usefulness of existing Erlang grade of service formulas, especially when faced with the evolving extensive changes in telephone networks. The old Markov traffic arrival and Markov call duration (or service) models are combined with what may be called more realistic user behavior and network scenarios. For the specified service system models, a generalized merger of ideal loss and ideal delay steady-state representations is developed. It is shown that a particular weighted inverse combination of Erlang B and C formulas, herein called the hybrid, is useful for grade of service estimates under several conditions. In addition to the usual offered load and the number of servers, the hybrid involves a relative Erlang B versus C weighting parameter θ. This parameter is non-negative and depends on system applications, as illustrated in a number of examples given. When less than or equal to unity, such as for the popular Poisson model, θ has a simple meaning. It is then the conditional probability that a typical customer elects to wait, or a service request is entered into a queue, given that all servers are busy.
18 citations
TL;DR: The queue network studied consists of n infinite queues in parallel served by independent servers and by other servers all linked to form a hierarchical structure that turns out to be useful and accurate with wide applicability.
Abstract: The queue network studied consists of n infinite queues in parallel served by independent servers and by other servers all linked to form a hierarchical structure. The total service a unit receives depends partially on other units in service. We call this type of servicing partially shared servicing. All interarrival times as well as service times are assumed exponentially distributed. The characteristic of interest is the traffic intensity of the infinite queues. Some simple formulae are obtained. An application to modelling a disc I/O system is described. The model turns out to be useful and accurate with wide applicability.
11 citations
Proceedings Article•
06 Feb 1979
5 citations
TL;DR: The conditional probability that the K + 1st server chosen is busy is shown to be equal to the a priori probability that a randomly chosen server in an equilibrium M/G/N-K-loss system with the same offered load is busy.
Abstract: If K + 1 servers in an equilibrium M/G/N-loss system are chosen randomly without replacement and the first K servers chosen are busy, the conditional probability that the K + 1st server chosen is busy is shown to be equal to the a priori probability that a randomly chosen server in an equilibrium M/G/N-K-loss system with the same offered load is busy.
TL;DR: The model presented is a generalization of some other models of queuing networks, and provides a more general definition of steady-state behavior, and a theoretical solution is presented.
Abstract: A periodic network is a queuing network whose steady-state behavior is not constant in time, but repeats itself in a cycle. This behavior may be caused by the introduction of periodic servers, e.g., paging drums. The model presented is a generalization of some other models of queuing networks, and provides a more general definition of steady-state behavior. A theoretical solution is presented. Examples of theoretical and approximate solutions are presented for a well-known queuing network model of a computer system.
TL;DR: The problem considered here is how to allocate n exponential tasks to two available servers, and arrivals are assumed to be Poisson, which gives rise to a network of queues with Erlang service times.
Abstract: The problem considered here is how to allocate n exponential tasks to two available servers. The servers may be placed in a series or parallel configuration, and arrivals are assumed to be Poisson. In general, this gives rise to a network of queues with Erlang service times. Alternative designs are examined using various approximation methods and simulation for several values of n and different utilization rates. If the workers must be placed in tandem, the rule of balancing the work of the two stations is found to be optimal for all cases considered. If it is possible to place the servers in a parallel configuration, each performing all n tasks, the average total waiting time can be substantially reduced.
01 Jan 1979
01 Jan 1979
TL;DR: The authors saw in chapter II that the analysis of a system consisti of an input server, a finite storage, and one other server was tractable; one could evaluate essentially anything one wishes, and most things that are of any practical significance can be evaluated without much difficulty.
Abstract: We saw in chapter II that the analysis of a system consisti of an input server, a finite storage, and one other server was tractable; one could evaluate essentially anything one wishes, and most things that are of any practical significance can be evaluated without much difficulty. As one adds additional servers and storages to the system, however, a complete analysis immediately becomes prohibitively difficult. One may devise schemes for evaluating special cases analytically (usually involving some infinite storages, some regular servers, i.e., ∆j = 0, or identical servers), or one can numerically evaluate, possibly by simulation, the performance of any specific system. To understand in detail, how the behavior of the system depends upon the ∆.∙s, μ.∙s, and cj∙s, however, is hopeless. Even if one could find an exact solution of the diffusion equation, the number of parameters in the solution would be so large that it would be impossible to comprehend the effects of each.
TL;DR: The performance of large scale multicell mobile radio systems using dynamic channel assignment and having limitations on the number of servers available in each coverage cell was investigated by computer simulation.
Abstract: The performance of large scale multicell mobile radio systems using dynamic channel assignment and having limitations on the number of servers available in each coverage cell was investigated by computer simulation. At low system blocking the traffic carried is determined solely by the server limitations while at higher blocking the traffic carried is determined by channel limitations.