Managing variances in manufacturing system design

The goal of this paper is to investigate how uncertainties in demand and production should be incorporated into manufacturing system design problems. We examine two problems in manufacturing system design: the resource allocation problem and the product grouping problem. In the resource allocation problem, we consider the issue of how to cope with uncertainties when we utilize two types of resources: actual processing capacity and stored capacity (inventory). A closed form solution of the optimal allocation scheme for each type of capacity is developed, and its performance is compared to that of the conventional scheme where capacity allocation and inventory control decisions are made sequentially. In the product grouping problem, we consider the issue of how we design production lines when each line is dedicated to a certain set of products. We formulate a mathematical program in which we simultaneously determine the number of production lines and the composition of each line. Two heuristics are developed for the problem.     

Inventory control in multi-channel retail

In recent years multi-channel retail systems have received increasing interest. Partly due to growing online business that serves as a second sales channel for many firms, offering channel specific prices has become a common form of revenue management. We analyze conditions for known inventory control policies to be optimal in presence of two different sales channels. We propose a single item lost sales model with a lead time of zero, periodic review and nonlinear non-stationary cost components without rationing to realistically represent a typical web-based retail scenario. We analyze three variants of the model with different arrival processes: demand not following any particular distribution, Poisson distributed demand and a batch arrival process where demand follows a Pòlya frequency type distribution. We show that without further assumptions on the arrival process, relatively strict conditions must be imposed on the penalty cost in order to achieve optimality of the base stock policy. We also show that for a Poisson arrival process with fixed ordering costs the model with two sales channels can be transformed into the well known model with a single channel where mild conditions yield optimality of an (s, S) policy. Conditions for optimality of the base stock and (s, S) policy for the batch arrival process with and without fixed ordering costs, respectively, are presented together with a proof that the batch arrival process provides valid upper and lower bounds for the optimal value function.     

Structural properties of the optimal resource allocation policy for single-queue systems

This paper studies structural properties of the optimal resource allocation policy for single-queue systems. Jobs arrive at a service facility and are sent one by one to a pool of computing resources for parallel processing. The facility poses a constraint on the maximum expected sojourn time of a job. A central decision maker allocates the servers dynamically to the facility. We consider two models: a limited resource allocation model, where the allocation of resources can only be changed at the start of a new service, and a fully flexible allocation model, where the allocation of resources can also change during a service period. In these two models, the objective is to minimize the average utilization costs whilst satisfying the time constraint. To this end, we cast these optimization problems as Markov decision problems and derive structural properties of the relative value function. We show via dynamic programming that (1) the optimal allocation policy has a work-conservation property, and (2) the optimal number of servers follows a step function with as extreme policy the bang-bang control policy. Moreover, (3) we provide conditions under which the bang-bang control policy takes place. These properties give a full characterization of the optimal policy, which are illustrated by numerical experiments.     

Optimal lateral transshipment policies for a two location inventory problem with multiple demand classes

We consider an inventory model for spare parts with two stockpoints, providing repairable parts for a critical component of advanced technical systems. As downtime costs for these systems are expensive, ready–for–use spare parts are kept in stock to be able to quickly respond to a breakdown of a system. We allow for lateral transshipments of parts between the stockpoints upon a demand arrival. Each stockpoint faces demands from multiple demand classes. We are interested in the optimal lateral transshipment policy. There are three ways in which a demand can by satisfied: from own stock, via a lateral transshipment, or via an emergency procedure. Using stochastic dynamic programming, we characterize and prove the structure of the optimal policy, that is, the policy for satisfying the demands which minimizes the average operating costs of the system. This optimal policy is a threshold type policy, with state-dependent thresholds at each stockpoint for every demand class. We show a partial ordering in these thresholds in the demand classes. In addition, we derive conditions under which the so-called hold back and complete pooling policies are optimal, two policies that are often assumed in the literature. Furthermore, we study several model extensions which fit in the same modeling framework.     

Capacity and Production Managment in a Single Product Manufacturing System

In planning and managing production systems, manufacturers have two main strategies for responding to uncertainty: they build inventory to hedge against periods in which the production capacity is not sufficient to satisfy demand, or they temporarily increase the production capacity by “purchasing” extra capacity. We consider the problem of minimizing the long-run average cost of holding inventory and/or purchasing extra capacity for a single facility producing a single part-type and assume that the driving uncertainty is demand fluctuation. We show that the optimal production policy is of a hedging point policy type where two hedging levels are associated with each discrete state of the system: a positive hedging level (inventory target) and a negative one (backlog level below which extra capacity should be purchased). We establish some ordering of the hedging levels, derive equations satisfied by the steady-state probability distribution of the inventory/backlog, and give a more detailed analysis of the optimal control policy in a two state (high and low demand rate) model.     

Fair resource allocation for demands with sharp lower tail inequalities

We consider a fairness problem in resource allocation where multiple groups demand resources from a common source with the total fixed amount. We show that for many common demand distributions satisfying sharp lower-tail inequalities, a natural allocation that provides resources proportional to each group's average demand performs very well. More specifically, this natural allocation is approximately fair and efficient (i.e., it provides near maximum utilization). We also show that, when a small amount of unfairness is allowed, the Price of Fairness (PoF), in this case, is close to 1.     

Controlling the GI/M/1 queue by conditional acceptance of customers

In this paper we consider the problem of controlling the arrival of customers into a GI/M/1 service station. It is known that when the decisions controlling the system are made only at arrival epochs, the optimal acceptance strategy is of a control-limit type, i.e., an arrival is accepted if and only if fewer than n customers are present in the system. The question is whether exercising conditional acceptance can further increase the expected long run average profit of a firm which operates the system. To reveal the relevance of conditional acceptance we consider an extension of the control-limit rule in which the nth customer is conditionally admitted to the queue. This customer may later be rejected if neither service completion nor arrival has occurred within a given time period since the last arrival epoch. We model the system as a semi-Markov decision process, and develop conditions under which such a policy is preferable to the simple control-limit rule.     

Cost-based decision-making in middleware virtualization environments

Middleware virtualization refers to the process of running applications on a set of resources (e.g., databases, application servers, other transactional service resources) such that the resource-to-application binding can be changed dynamically on the basis of applications’ resource requirements. Although virtualization is a rapidly growing area, little formal academic or industrial research provides guidelines for cost-optimal allocation strategies. In this work, we study this problem formally. We identify the problem and describe why existing schemes cannot be applied directly. We then formulate a mathematical model describing the business costs of virtualization. We develop runtime models of virtualization decision-making paradigms. We describe the cost implications of various runtime models and consider the cost effects of different managerial decisions and business factors, such as budget changes and changes in demand. Our results yield useful insights for managers in making virtualization decisions.     

Optimal resource allocation across related channels

In this paper, we consider resource allocation strategies of a limited resource across two related channels in a multi-period setting. We study a stochastic control problem where the objective is to determine the optimal limited resource allocation policy across two related channels and optimal transshipment policy between these two channels. We characterize some structural results of the optimal resource allocation policy and show that it is determined by three monotone curves.     

Managing an integrated production inventory system with information on the production and demand status and multiple non-unitary demand classes

In this paper, we study a system consisting of a manufacturer or supplier serving several retailers or clients. The manufacturer produces a standard product in a make-to-stock fashion in anticipation of orders emanating from n retailers with different contractual agreements hence ranked/prioritized according to their importance. Orders from the retailers are non-unitary and have sizes that follow a discrete distribution. The total production time is assumed to follow a k₀-Erlang distribution. Order inter-arrival time for class l demand is assumed to follow a k_l-Erlang distribution. Work-in-process as well as the finished product incur a, per unit per unit of time, carrying cost. Unsatisfied units from an order from a particular demand class are assumed lost and incur a class specific lost sale cost. The objective is to determine the optimal production and inventory allocation policies so as to minimize the expected total (discounted or average) cost. We formulate the problem as a Markov decision process and show that the optimal production policy is of the base-stock type with base-stock levels non-decreasing in the demand stages. We also show that the optimal inventory allocation policy is a rationing policy with rationing levels non-decreasing in the demand stages. We also study several important special cases and provide, through numerical experiments, managerial insights including the effect of the different sources of variability on the operating cost and the benefits of such contracts as Vendor Managed Inventory or Collaborative Planning, Forecasting, and Replenishment. Also, we show that a heuristic that ignores the dependence of the base-stock and rationing levels on the demands stages can perform very poorly compared to the optimal policy.     

Supply chain inventory optimization with two customer classes in discrete time

In this paper we consider a single-item inventory system where two demand classes with different service requirements are satisfied from a common inventory. A critical level, reorder point, order quantity or (s, q, k) policy is in use. The time axis is divided into discrete time units, which is a common characteristic of many real-life supply-chain processes. The inventory process within the lead time of a replenishment order is modelled as a sequence of (1) an ordinary renewal process and (2) two alternating renewal processes. Approximations are developed for the demand class-specific fill rates and the probability distribution of the waiting time of low priority customer orders. This waiting time distribution is used for the inventory allocation in a two-stage supply chain.     

Two-resource stochastic capacity planning employing a Bayesian methodology

We examine a stochastic capacity-planning problem with two resources that can satisfy demand for two services. One of the resources can only satisfy demand for a specific service, whereas the other resource can provide both services. We formulate the problem of choosing the capacity levels of each resource to maximize expected profits. In addition, we provide analytic, easy-to-interpret optimal solutions, as well as perform a comparative statics analysis. As applying the optimal solutions effectively requires good estimates of the unknown demand parameters, we also examine Bayesian estimates of the demand parameters derived via a class of conjugate priors. We compare the optimal expected profits when demands for the two services follow independent distributions with informative and non-informative priors, and demonstrate that using good informative priors on demand can significantly improve performance.     

A global optimization algorithm for reliable network design

In this paper, we consider the problem of designing reliable networks that satisfy supply/demand, flow balance, and capacity constraints, while simultaneously allocating certain resources to mitigate the arc failure probabilities in such a manner as to minimize the total cost of network design and resource allocation. The resulting model formulation is a nonconvex mixed-integer 0-1 program, for which a tight linear programming relaxation is derived using RLT-based variable substitution strategies and a polyhedral outer-approximation technique. This LP relaxation is subsequently embedded within a specialized branch-and-bound procedure, and the proposed approach is proven to converge to a global optimum. Various alternative partitioning strategies that could potentially be employed in the context of this branch-and-bound framework, while preserving the theoretical convergence property, are also explored. Computational results are reported for a hypothetical scenario based on different parameter inputs and alternative branching strategies. Related optimization models that conform to the same class of problems are also briefly presented.     

Dynamic resource allocation in a multi-product make-to-stock production system

We consider optimal policies for a production facility in which several (K) products are made to stock in order to satisfy exogenous demand for each. The single machine version of this problem in which the facility manufactures at most one product at a time to minimise inventory costs has been much studied. We achieve a major generalisation by formulating the production problem as one involving dynamic allocation of a key resource which drives the manufacture of all products under an assumption that each additional unit of resource allocated to a product achieves a diminishing return of increased production rate. A Lagrangian relaxation of the production problem induces a decomposition into K single product problems in which the production rate may be varied but is subject to charge. These reduced problems are of interest in their own right. Under mild conditions of full indexability the Lagrangian relaxation is solved by a production policy with simple index-like structure. This in turn suggests a natural index heuristic for the original production problem which performs strongly in a numerical study. The paper discusses the importance of full indexability and makes proposals for the construction of production policies involving resource idling when it fails.     

Optimal integrated production and inventory control of an assemble-to-order system with multiple non-unitary demand classes

We study a pure assemble-to-order system subject to multiple demand classes where customer orders arrive according to a compound Poisson process. The finished product is assembled from m different components that are produced on m distinct production facilities in a make-to-stock fashion. We show that the optimal production policy of each component is a state-dependent base-stock policy and the optimal inventory allocation policy is a multi-level state-dependent rationing policy. Using numerical experimentation, we first study the system behavior as a function of order size variability and order size. We show that the optimal average cost rate is more sensitive to order size variability than to order size. We also compare the optimal policy to the first-come first-serve policy and show that there is great benefit to inventory rationing. We also propose two simple heuristics and show that these can effectively mimic the optimal policy which is generally much more difficult to determine and, especially, to implement.     

Towards an Erlang formula for multiclass networks

Consider a multiclass stochastic network with state-dependent service rates and arrival rates describing bandwidth-sharing mechanisms as well as admission control and/or load balancing schemes. Given Poisson arrival and exponential service requirements, the number of customers in the network evolves as a multi-dimensional birth-and-death process on a finite subset of ℕ ^k . We assume that the death (i.e., service) rates and the birth rates depending on the whole state of the system satisfy a local balance condition. This makes the resulting network a Whittle network, and the stochastic process describing the state of the network is reversible with an explicit stationary distribution that is in fact insensitive to the service time distribution. Given routing constraints, we are interested in the optimal performance of such networks. We first construct bounds for generic performance criteria that can be evaluated using recursive procedures, these bounds being attained in the case of a unique arrival process. We then study the case of several arrival processes, focusing in particular on the instance with admission control only. Building on convexity properties, we characterize the optimal policy, and give criteria on the service rates for which our bounds are again attained.     

Dynamic Scheduling of a Multiclass Fluid Model with Transient Overload

We study the optimal dynamic scheduling of different requests of service in a multiclass stochastic fluid model that is motivated by recent and emerging computing paradigms for Internet services and applications. In particular, our focus is on environments with specific performance guarantees for each class under a profit model in which revenues are gained when performance guarantees are satisfied and penalties are incurred otherwise. Within the context of the corresponding fluid model, we investigate the dynamic scheduling of different classes of service under conditions where the workload of certain classes may be overloaded for a transient period of time. Specifically, we consider the case with two fluid classes and a single server whose capacity can be shared arbitrarily among the two classes. We assume that the class 1 arrival rate varies with time and the class 1 fluid can more efficiently reduce the holding cost. Under these assumptions, we characterize the optimal server allocation policy that minimizes the holding cost in the fluid model when the arrival rate function for class 1 is known. Using the insights gained from this deterministic case, we study the stochastic fluid system when the arrival rate function for class 1 is random and develop various policies that are optimal or near optimal under various conditions. In particular, we consider two different types of heavy traffic regimes and prove that our proposed policies are strongly asymptotically optimal. Numerical examples are also provided to demonstrate further that these policies yield good results in terms of minimizing the expected holding cost.     

基于MAS协作机制的动态计算资源调度模型

针对网格环境的自治性、动态性、分布性和异构性等特征.提出基于多智能体系统(Mutil Agent System,MAS)博弈协作的资源动态分配和任务调度模型,建立了能够反映供求关系的网格资源调度模型和任务求解算法,证明了资源分配博弈中Nash均衡点的存在性、唯一性和Nash均衡解,该方法能够利用消费者agent的学习和协商能力,考虑和引入消费者的心理行为,使得消费者的资源申请和任务调度具有较高的合理性和有效性.实验结果表明,资源调度算法不但可以有效减少不必要的延迟,而且在响应时间的平滑性、吞吐率及资源利用率方面比传统算法要好,从而使得整个资源的供需合理、负载均衡.     

具有凸资源约束的单机指数学习效应问题

研究在所有工件的正常加工时间均相同的情况下具有指数学习效应和凸资源约束的单机排序问题.给出了两种模型:在资源消耗总费用有限的情况下,以工件的最大完工时间为目标函数;在工件的最大完工时间有限的情况下,以资源消耗总费用为目标函数.求两种模型下的最优排序和最优资源分配,使得目标函数最小.证明这两个问题都是多项式时间可解的,并给出了相应的算法.     

Multi-item inventory control with full truckloads: A comparison of aggregate and individual order triggering

In this paper, we consider the stochastic joint replenishment problem in an environment where transportation costs are dominant and full truckloads or full container loads are required. One replenishment policy, taking into account capacity restrictions of the total order volume, is the so-called QS policy, where replenishment orders are placed to raise the individual inventory positions of all items to their order-up-to levels, whenever the aggregate inventory position drops below the reorder level. We first provide a method to compute the policy parameters of a QS policy such that item target service levels can be met, under the assumption that demand can be modeled as a compound renewal process. The approximation formulas are based on renewal theory and are tested in a simulation study which reveals good performance. Second, we compare the QS policy with a simple allocation policy where replenishment orders are triggered by the individual inventory positions of the items. At the moment when an individual inventory position drops below its item reorder level, a replenishment order is triggered and the total vehicle capacity is allocated to all items such that the expected elapsed time before the next replenishment order is maximized. In an extensive simulation study it is illustrated that the QS policy outperforms this allocation policy since it results in lower inventory levels for the same service level. Although both policies lead to similar performance if items are identical, it can differ substantially if the item characteristics vary.