Performance analysis of a hybrid MPI/OpenMP application on multi-core clusters

The mixing of shared memory and message passing programming models within a single application has often been suggested as a method for improving scientific application performance on clusters of shared memory or multi-core systems. DL_POLY, a large scale molecular dynamics application programmed using message passing programming has been modified to add a layer of shared memory threading and the performance analysed on two multi-core clusters. At lower processor numbers, the extra overheads from shared memory threading in the hybrid code outweigh performance benefits gained over the pure MPI code. On larger core counts the hybrid model performs better than pure MPI, with reduced communication time decreasing the overall runtime.     

A cross entropy approach to design of reliable networks

One of the most important parameters determining the performance of communication networks is network reliability. The network reliability strongly depends on not only topological layout of the communication networks but also reliability and availability of the communication facilities. The selection of optimal network topology is an NP-hard problem so that computation time of enumeration-based methods grows exponentially with network size. This paper presents a new solution approach based on cross-entropy method, called NCE, to design of communication networks. The design problem is to find a network topology with minimum cost such that all-terminal reliability is not less than a given level of reliability. To investigate the effectiveness of the proposed NCE, comparisons with other heuristic approaches given in the literature for the design problem are carried out in a three-stage experimental study. Computational results show that NCE is an effective heuristic approach to design of reliable networks.     

Adaptive Gaussian process emulators for efficient reliability analysis

This paper presents an approximation method for performing efficient reliability analysis with complex computer models. The computational cost of industrial-scale models can cause problems when performing sampling-based reliability analysis. This is due to the fact that the failure modes of the system typically occupy a small region of the performance space and thus require relatively large sample sizes to accurately estimate their characteristics. The sequential sampling method proposed in this article, combines Gaussian process-based optimisation and subset simulation. Gaussian process emulators construct a statistical approximation to the output of the original code, which is both affordable to use and has its own measure of predictive uncertainty. Subset simulation is used as an integral part of the algorithm to efficiently populate those regions of the surrogate which are likely to lead to the performance function exceeding a predefined critical threshold. The emulator itself is used to inform decisions about efficiently using the original code to augment its predictions. The iterative nature of the method ensures that an arbitrarily accurate approximation of the failure region is developed at a reasonable computational cost. The presented method is applied to an industrial model of a biodiesel filter.     

A consistent implementation for inelastic materials in an ALE formulation for steady state rolling contact

The numerical analysis of rolling contact for rubber materials is a challenging task, especially due to the many nonlinearities inherent to the material, large deformations, friction, and energy dissipation, among others. Industrial applications can be found in ball bearings, rollers, and most commonly in tires of vehicles, applications where reliable numerical simulations lead to the improvement of durability, performance and safety. While a transient analysis stands as a practical and powerful tool for the simulation of rotating bodies, the large amount of computational resources required represents its biggest disadvantage. An alternative frequently used lays in a steady state simulation by means of an Arbitrary Lagrangian Eulerian (ALE) formulation, where the rotational velocity and axial loads are assumed to remain constant. Within this framework, the reference configuration is neither attached to the material particles nor fixed in space and special attention should be paid to the history variables of inelastic materials. In this work, a viscoelastic material model is implemented in an in-house finite element code, based on a generalized Maxwell model. The implementation takes into consideration the contribution of all elements connected in circumferential direction and a consistent linearization is made for each of them, leading to an assembled stiffness matrix with more non-zero values than a standard one. This approach is combined with smeared reinforcement embedded in base elements. The reinforcing layers are described by a hyperelastic material model, providing additional advantages for the modeling and simulation of reinforced rollers and tires. Numerical results for different examples show the capabilities of this implementation and the efficiency of the numerical algorithms is discussed. Important remarks and an outlook for further research concludes this presentation. (© 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim)     

Randomized strategy equilibrium in the action commitment game with costs of leading

We investigate a two-player action commitment game where one simultaneous-move and two sequential-move pure strategy equilibria exist when the cost of leading is zero, while the simultaneous-move outcome is not an equilibrium when the leading cost is small positive. We show that this discontinuity disappears if we consider randomized strategy equilibria. We investigate a price competition model and show that randomized strategy equilibria exist and any of them converges to the Bertrand equilibrium when the leading cost converges to zero.     

Routeing in a Network with Multi-Class Links

Consider a communication network with certain nodes and different types of links. In addition to the normal link cost, a transformation cost is charged if the incoming link and the outgoing link are of different types. An optimal routeing from a given node to its destination node is sought. The major difficulty in handling this problem is that the principle of optimality does not hold. A model with node separation is built to overcome this difficulty. By using the new model, the original routeing problem is no more than a shortest-path problem. Hence, we can implement this model to current electronic switching machines.     

The offshore wind farm array cable layout problem: a planar open vehicle routing problem

In an offshore wind farm (OWF), the turbines are connected to a transformer by cable routes that cannot cross each other. Finding the minimum cost array cable layout thus amounts to a vehicle routing problem with the additional constraints that the routes must be embedded in the plane. For this problem, both exact and heuristic methods are of interest. We optimize cable layouts for real-world OWFs by a hop-indexed integer programming formulation, and develop a heuristic for computing layouts based on the Clarke and Wright savings heuristic for vehicle routing. Our heuristic computes layouts on average only 2% more expensive than the optimal layout. Finally, we present two problem extensions arising from real-world OWF cable layouts, and adapt the integer programming formulation to one of them. The thus obtained optimal layouts are up to 13% cheaper than the actually installed layouts.     

Detailed layout planning for irregularly-shaped machines with transportation path design

In order to obtain a competitive level of productivity in a manufacturing system, efficient machine or department arrangements and appropriate transportation path structures are of considerable importance. By defining a production system’s basic structure and material flows, the layout determines its operational performance over the long term. However, most approaches proposed in the literature provide only a block layout, which neglects important operational details. By contrast, in this paper, we introduce approaches to planning layouts at a more detailed level. Hence, this present paper introduces an integrated approach which allows a more detailed layout planning by simultaneously determining machine arrangement and transportation paths. Facilities to be arranged as well as the entire layout may have irregular shapes and sizes. By assigning specific attributes to certain layout subareas, application-dependent barriers within the layout, like existing walls or columns, can be incorporated. We introduce a new mathematical layout model and develop several improvement procedures. An analysis of the computational experiments shows that more elaborate heuristics using variable neighborhoods can generate promising layout configurations.     

Timing market entry with a contribution-maximization approach to location-allocation decisions

This paper develops a multiperiod model which considers revenue and cost effects of location-allocation decisions. The dynamic model is formulated as a zero-one linear program and is interpreted as a vehicle for analyzing the timing of entry into markets.The paper examines an important special case of the model which applies under the assumptions of non-negative contributions from all potential allocations and monotonically non-decreasing demands over time. For this special case, an exact algorithm is developed and is shown to be substantially more efficient than a general-purpose code for zero-one problems. The new algorithm is capable of solving practical-sized versions of the model with moderate computational requirements.     

A Parallel Code for LES of Compressible Swirling Jet Flow Undergoing Vortex Breakdown

Swirling jets undergoing vortex breakdown occur in many technical applications, e.g. vortex burners, turbines and jet engines. To simulate the highly nonlinear dynamics of the flow, it is necessary to use high-order numerical methods, leading to increased computational cost. To be able to perform simulations in acceptable turn-around time, an available LES code for solving the filtered compressible Navier-Stokes equations in cylindrical coordinates using compact finite-difference schemes was parallelized for massively-parallel architectures. The parallelization was done following the ghost-cell approach for filtering in the three spatial directions. The inter-process communication is handled using the message passing interface (MPI). The weak and strong scaling properties of the code indicate that it can be used for massively parallel simulations using several thousand processors. (© 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim)     

Cache placement in sensor networks under an update cost constraint

In this paper, we address an optimization problem that arises in the context of cache placement in sensor networks. In particular, we consider the cache placement problem where the goal is to determine a set of nodes in the network to cache/store the given data item, such that the overall communication cost incurred in accessing the item is minimized, under the constraint that the total communication cost in updating the selected caches is less than a given constant. In our network model, there is a single server (containing the original copy of the data item) and multiple client nodes (that wish to access the data item). For various settings of the problem, we design optimal, near-optimal, heuristic-based, and distributed algorithms, and evaluate their performance through simulations on randomly generated sensor networks.     

考虑回收-补偿约束的卫星城镇生活垃圾中转站选址研究

为进行卫星城镇WSW中转站优化选址,本文结合城镇中转站选址经济、社会、环境负效应等相关影响因素细化分析,确定了城镇中转设施分步优化模型方法。该方法结合规范约束距离改进LSCP模型,于待选点中确定一次覆盖所有收集点的设施备选组合,而后应用综合考虑了建设运营成本、负效应、回收效率的多目标整数规划模型最终确定中转站位置、收纳量与服务分配方案。通过实例分析,验证了本文方法实用性与合理性。     

合同能源管理中超额节能收益分配问题研究

在我国合同能源管理项目中,节能收益分享模式被广泛采用。通过构建“委托-代理”模型深入分析节能收益分享模式下的超额节能收益分配问题。研究发现:(1)无论是否出现超额节能收益,合同中用能方的节能收益分享比例不得超过50%;(2)在引入节能服务公司的努力成本系数且不考虑节能改造双方风险偏好的情况下(风险中性),当努力成本系数在特定的取值范围内,优化后的超额节能收益分配方式对节能服务公司更具激励作用;(3)考虑节能服务公司为风险规避时,其最优期望节能收益不但受其努力成本系数的影响,还受其风险规避程度以及节能收益期望方差的影响。     

Performance comparisons for maintained items

Most maintenance studies deal with the problem of maintenance-costs optimization, given the cost of the available maintenance actions and a cost objective function. The focus in this paper is on yet another relevant aspect namely the improvement in performance that results from a maintenance action in terms of performance criteria of interest. The evaluation of performance improvement requires comparisons of random quantities which is done by means of suitable probabilistic ordering notions. Different modelling situations are considered, and for each of them conditions are obtained on the life distributions of the present item, and the new one if of different type, that ensure performance improvement. Apart from providing a decision tool for maintenance application, the paper also contributes to clarifying relationships between (probabilistic) ordering notions and aging properties within the maintenance context.     

A model for supply chain design considering the cost of quality

Recent studies have shown that the cost of quality (COQ) is of more strategic and economic importance than previously conceived. Whereas previous works have applied COQ as an internal performance measure within companies, the purpose of this paper is to present a model for supply chain design that computes the COQ as a global performance measure for the entire supply chain. In addition, rather than assume an exogenously given COQ curve, our model computes COQ in terms of internal operational decisions such as the error rate at inspection and fraction defective at manufacturing. The model can be used to design a logistic route that achieves a minimum total cost while maintaining an overall quality level and to evaluate the impact of investment in quality to increase overall profits. The behaviour of the model is illustrated with numerical examples that show how the COQ function changes depending on various parameters.     

Flexible hiring in a make to order system with parallel processing units

In this paper, we study a make-to-order production system with parallel, identical processing units. Each order needs to be satisfied on a single processing unit that is run by a crew. The inter-arrival time and the service time for each order are random variables. The system operates under a lead time performance constraint, which demands the completion of each order within a pre-determined lead time with a certain probability. The minimum number of processing units needed to satisfy this constraint is determined at the tactical level. Our research focuses on the cost savings that can be realized with the use of flexible crews via contractual hiring agreements with an External Labor Supply Agency (ELSA). The ELSA can periodically provide an agreed number of crews. The cost incurred for a flexible crew is higher than that for a permanent crew, and is decreasing in the period length. We model and analyze this system using the transient behavior analysis of multi-server queues and propose several empirically testable functions for the cost of flexible crews. In our computational study, we demonstrate possible cost savings of 2-level, threshold type hiring policies, relative to the fixed capacity system, under 9 scenarios with three demand-to-processing rate ratios and three lead time performance constraints, each of which reflects a different level of ambition. We observe that the maximum savings occur when the cost of a flexible crew is same as that of a permanent crew, and range from 29.38% to 50.56%. However, as the flexible crews become more expensive, the system may choose to employ permanent crews only. We observe that cost savings consist of two parts: savings due to the cancellation of the sclerosis of capacity discreteness, and savings due to the use of workload information in hiring actions. The latter part is higher for more ambitious lead time performance constraints, and for higher mean processing times. Finally, when there is an additional cost for transacting an agreement with the ELSA, we observe that the capacity flexibility option loses its charm, especially if the transaction cost is higher than the cost of a permanent crew.     

Designing Least-Cost Nonblocking Broadband Networks

Integrated network technologies, such as ATM, support multimedia applications with vastly different bandwidth needs, connection request rates, and holding patterns. Due to their high level of flexibility and communication rates approaching several gigabits per second, the classical network planning techniques, which rely heavily on statistical analysis, are less relevant to this new generation of networks. In this paper, we propose a new model for broadband networks and investigate the question of their optimal topology from a worst-case performance point of view. Our model is more flexible and realistic than others in the literature, and our worst-case bounds are among the first in this area. Our results include a proof of intractability for some simple versions of the network design problem and efficient approximation algorithms for designing nonblocking networks of provably small cost. More specifically, assuming some mild global traffic constraints, we show that a minimum-cost nonblockingstarnetwork achieves near-optimal cost; the cost ratio is at most 2 if switch source and sink capacities are symmetric and at most 3 when the total source and sink capacities are balanced. In the special case of unit link costs, we can show that a star network is indeed the cheapest nonblocking network.     

A formal framework to analyze cost and performance in Map-Reduce based applications

Satisfying the global throughput targets of scientific applications is an important challenge in high performance computing (HPC) systems. The main difficulty lies in the high number of parameters having an important impact on the overall system performance. These include the number of storage servers, features of communication links, and the number of CPU cores per node, among many others.In this paper we present a model that computes a performance/cost ratio using different hardware configurations and focusing on scientific computing. The main goal of this approach is to balance the trade-off between cost and performance using different combinations of components for building the entire system. The main advantage of our approach is that we simulate different configurations in a complex simulation platform. Therefore, it is not necessary to make an investment until the system computes the different alternatives and the best solutions are suggested. In order to achieve this goal, both the system's architecture and Map-Reduce applications are modeled. The proposed model has been evaluated by building complex systems in a simulated environment using the SIMCAN simulation platform.     

Supply chain models for an assembly system with preprocessing of raw materials

In this paper, we investigate the material procurement and delivery policy in a production system where raw materials enter into the assembly line from two different flow channels. The system encompasses batch production process in which the finished product demand is approximately constant for an infinite planning horizon. Two distinct types of raw materials are passed through the assembly line before to convert them into the finished product. Of the two types of raw materials, one type requires preprocessing inside the facility before the assembly operation and other group is fed straightway in the assembly line. The conversion factors are assigned to raw materials to quantify the raw material batch size required. To analyze such a system, we formulate a nonlinear cost function to aggregate all the costs of the inventories, ordering, shipping and deliveries. An algorithm using the branch and bound concept is provided to find the best integer values of the optimal solutions. The result shows that the optimal procurement and delivery policy minimizes the expected total cost of the model. Using a test problem, the inventory requirements at each stage of production and their corresponding costs are calculated. From the analysis, it is shown that the rate and direction change of total cost is turned to positive when delivery rates per batch reaches close to the optimal value and the minimum cost is achieved at the optimal delivery rate. Also, it is shown that total incremental cost is monotonically increasing, if the finished product batch size is increased, and if, inventory cost rates are increased. We examine a set of numerical examples that reveal the insights into the procurement-delivery policy and the performance of such an assembly type inventory model.     

A comparison of average and discounted cost models for the dynamic lot size inventory problem

It is a common practice in the inventory literature to use average cost models as approximations to the theoretically correct discounted cost models. An average cost model minimizes the average undiscounted cost per period, while a discounted cost model minimizes the total discounted cost over the problem horizon. This paper attempts to answer an important question: How good are the results (the total discounted costs) for the average cost models compared to those for the discounted cost models? This question has been conclusively answered for the simplest inventory model where the demand rate and other parameters are assumed to remain constant in time. This paper addresses this issue for the first time for the case where demand rates are allowed to be nonstationary in time.A discounted cost model has been developed in the paper to carry out this comparison. It is shown that a simple dynamic programming algorithm can be used to find optimal order policies for the discounted cost model.The effect of the varying interest rates and other parameters on the relative performance of the average cost model has been studied by developing an insightful analysis and also by doing a computational study. The results show that, while the average cost model can cost as much as about 26% more than the discounted cost model in extreme cases, this increase is not significant for the parameter values in the range of the common interest.