A matrix-geometric approximation for tandem queues with blocking and repeated attempts

Our interest is in the study of the MAP/PH/1/1→·/PH/1/K+1 queue with blocking and repeated attempts. The main feature of its infinitesimal generator is the spatial heterogeneity caused by the transitions due to successful repeated attempts. We develop an algorithmic solution by making a simplifying approximation which yields an infinitesimal generator which is spatially homogeneous and has a modified matrix-geometric stationary vector. The essential tool in our analysis is the general theory on quasi-birth-and-death processes.     

同步N—策略多重休假M／M／c排队

研究了具有同步N-策略多重休假的M／M／c排队系统．在休假时间服从相型（PH）分布的假设下，给出了系统的稳态指标．证明在已知服务台全忙并且系统中顾客数大于或等于N的条件下，条件随机变量可分解成独立随机变量之和，其中一个是无休假经典M／M／c系统中的对应条件变量，另一个是休假引起的附加随机变量     

关于两类M/M/1排队系统的渐进稳定性

讨论了两类M/M/1排队系统的关系,通过较简单的方法得到了系统的稳态解是渐进稳定的.     

Transient behaviour of a double-channel Markovian queue with limited waiting space

For a double channel Markovian queue with finite waiting space the difference equations satisfied by the Laplace transforms of the state probabilities at finite time are solved and the state probabilities are obtained in a simple closed form which can be easily used to find the important parameters of the system.     

Discrete-Time Geo X /G/1 Queue with Preemptive Repeat Different Priority

This paper considers a discrete-time Geo ^X /G/1 queue accepting two classes of messages with preemptive repeat different priority. Service times of messages of each priority class are i.i.d. according to a general discrete distribution function that may differ between two classes. The completion time and the stability condition for our system are investigated. By using the supplementary variable method and the generating function technique, we derive the joint system contents distributions at various observation instants and also compute the probability distribution for the unfinished work.     

Dial-Up Internet Access: A Two-Provider Cost Model

We present a cost model for splitting Internet dial-up traffic (which varies by time-of-day) between two large modem banks. One of the modem banks charges by the hour, the other charges for the peak number-in-system during the day. To study if the possible savings are enough to make the effort worthwhile, we formulate a clairvoyant (“perfect information”) Integer Program that is equivalent to a network flow problem. This leads us to use a ceiling policy. In the stochastic control case, we use a Modified Offered Load (MOL) approximation to explore the properties of the system, and develop a square-root-type rule to set the ceiling in the homogeneous case. We also use simulation to determine an optimal ceiling when we cannot route individual calls precisely. We propose approximations that may be computed for any call duration distribution, and compare their answers to exact differential-equation procedures for Exponential call durations. AMS subject classification: 60K25, 90B18, 68M20, 90B22, 60K30     

On the exact asymptotic behaviour of the distribution of ladder epochs

Let T₊ denote the first increasing ladder epoch in a random walk with a typical step-length X. It is known that for a large class of random walks with E(X)=0,E(X²)=∞, and the right-hand tail of the distribution function of X asymptotically larger than the left-hand tail, $\text{P}\text{T}{}_{\mathrm{+}}\text{?n∽n}{}^{\mathrm{<mtext>1</mtext><mtext>\beta ?1</mtext>}}\text{L}{}_{\mathrm{+}}\text{(n)}$ as n→∞, with 1<β<2 and L₊ slowly varying, if and only if$\text{P}\text{{X?x}∽ 1/{x}{}^{\mathrm{\beta }}\text{L(x)}}$ as x→+∞, with L slowly varying. In this paper it is shown how the asymptotic behaviour of L determines the asymptotic behaviour of L₊ and vice versa. As a by-product, it follows that a certain class of random walks which are in the domain of attraction of one-sided stable laws is such that the down-going ladder height distribution has finite mean.     

Single line queue with recurrent repeated demands

We analyze a model queueing system in which customers cannot be in continuous contact with the server, but must call in to request service. If the server is free, the customer enters service immediately, but if the server is occupied, the unsatisfied customer must break contact and reapply for service later. There are two types of customer present who may reapply. First transit customers who arrive from outside according to a Poisson process and if they find the server busy they join a source of unsatisfied customers, called the orbit, who according to an exponential distribution reapply for service till they find the server free and leave the system on completion of service. Secondly there are a number of recurrent customers present who reapply for service according to a different exponential distribution and immediately go back in to the orbit after each completion of service. We assume a general service time distribution and calculate several characterstic quantities of the system for both the constant rate of reapplying for service and for the case when customers are discouraged and reduce their rate of demand as more customers join the orbit.     

Tuning Garwick's algorithm for repacking sequential storage

Garwick's algorithm, for repacking LIFO lists stored in a contiguous block of memory, bases the allocation of remaining space upon both sharing and previous stack growth. A system whereby the weight applied to each method can be adjusted according to the current behavior of the stacks is discussed.We also investigate the problem of determining during memory repacking that the memory is used to saturation and the driving program should therefore be aborted. The tuning parameters studied here seem to offer no new grasp on this problem.     

Performance Analysis with Truncated Heavy-Tailed Distributions

This paper deals with queues and insurance risk processes where a generic service time, resp. generic claim, has the form U ∧ K for some r.v. U with distribution B which is heavy-tailed, say Pareto or Weibull, and a typically large K, say much larger than . We study the compound Poisson ruin probability ψ(u) or, equivalently, the tail of the M/G/1 steady-state waiting time W. In the first part of the paper, we present numerical values of ψ(u) for different values of K by using the classical Siegmund algorithm as well as a more recent algorithm designed for heavy-tailed claims/service times, and compare the results to different approximations of ψ(u) in order to figure out the threshold between the light-tailed regime and the heavy-tailed regime. In the second part, we investigate the asymptotics as K → ∞ of the asymptotic exponential decay rate γ = γ ^(K) in a more general truncated Lévy process setting, and give a discussion of some of the implications for the approximations. AMS 2000 Subject Classification Primary 68M20, Secondary 60K25 †Partially supported by MaPhySto—A Network in Mathematical Physics and Stochastics, founded by the Danish National Research Foundation. An erratum to this article is available at .