企业私有云环境下面向高性能计算的资源弹性分配算法

为应对企业私有云环境下高性能计算面临的提高集群资源利用率和保障用户计算能力的挑战,本文提出了一种基于用户资源配额的资源弹性分配算法(QREA)。该算法在计算资源充足时,通过将空闲资源追加给弹性作业以提高资源利用率;当计算资源紧张时,通过资源回收以保障用户的资源配额。为提高多用户共享集群资源的公平性,提出了基于用户历史信用值的资源追加和回收算法。大量的实验结果表明,在低作业负载下,QREA算法与基本资源分配算法BRA和多用户公平调度算法Fair Scheduler相比,作业平均完成时间缩短50%;在高作业负载下,QREA算法比BRA算法在资源利用率上提高10%,在多用户公平性上,比基于当前资源使用情况的资源弹性分配算法CREA提高34. 5%。     

面向混合负载的集群资源弹性调度

针对目前集群资源调度方法难以适应互联网业务多样化、定制化特征的问题,提出了一种面向混合负载的集群资源弹性调度方法.该方法通过构建作业约束描述语言,允许作业基于自身负载特征提出多维度的资源申请和具有负载意识的资源调度算法,实现在同一集群内各类业务统一部署与管理,及时匹配资源需求的变化;通过建立作业的软约束与硬约束之间的转化机制,满足作业在不同执行阶段对资源的定制化需求.实验表明,该方法相比于Hadoop,可允许作业利用较少资源获得更优性能,在实际生产系统中,基于该方法可将集群资源利用率由62％提升到75％.     

HDAS:异构集群上Hadoop+框架中的动态亲和性调度

针对现有异构集群的编程框架着重于异构资源的利用,没有充分考虑共享资源竞争导致作业完成时间延长的情况,基于Hadoop+框架和异构任务模型,提出并实现了异构动态亲和性调度(HDAS)算法,该算法利用Hadoop的心跳机制监测各结点上的资源使用情况和实时负载,对系统中的异构资源用不同的策略计算与任务的亲和性,进行任务分派,使系统的资源利用更充分,从而降低共享资源竞争导致的任务延迟,提高系统的整体吞吐率,且提交到系统中的应用都会在启动后一定时间内被执行。对25种混合负载的试验表明,Hadoop+框架使用HDAS相对于Hadoop的实现可获得平均21.9x的加速比,明显优于基于异构任务模型的调度策略(17.9x),并使其中21个负载的任务平均延迟不超过6%,在任务对系统资源需求多样性丰富的混合负载上优化效果明显。     

基于多重约束的飞机移动装配线作业调度优化

基于对飞机移动装配线作业多重约束的分析,建立了作业调度优化问题的整数规划模型,提出了组合优先规则WRST以及基于此规则的启发式算法.WRST综合考虑了最迟完成时间、资源以及空间稀缺度三个指标;而算法设计包括基于WRST规则的作业拓扑排序和作业资源分配两阶段.数值试验结果显示基于WRST规则的启发式算法求解质量高,求解速...     

合理利用测控资源的动态调度模型

通过研究小卫星系统测控需求，地面测控资源以及两者之间的关系，建立了一套测控资源动态调度数学模型，以反映小卫星测控要求和测控资源的函数关系，旨在寻找一种资源分配随时间动态改变的最优调度方法，本文提出的测控资源动态调度模型，可用于小卫星系统概念设计阶段运行费用的估算以及提高测控中心的自动化程度，从而保证小卫星大效益。     

基于工作流网的应急资源配置与路径规划集成优化

应急管理决策通常包括站点选址、资源配置、运输调度等内容,如何从应急处置整体流程控制的视角对决策内容进行集成建模及优化,是应急管理研究付诸实际应用的关键。本文提出具有资源和不确定时间约束的应急工作流网模型,通过三类库所（状态库所、动作库所、资源库所）及三类时间属性（可视时间、静态时间、动态时间）,揭示多部门联合应急中的作业时序与资源占用关系。在给定整体流程最大完成时间的条件下,以资源消耗与占用成本、资源运输与惩罚成本总和为目标函数,建立应急资源配置与路径规划的集成问题模型,并采用遗传粒子群混合算法对问题进行求解。根据遗传优化得到的应急资源配置方案,借助应急工作流网计算各动作库所、状态库所的时间参数,以此作为约束条件利用嵌套的粒子群算法进行资源运输策略优化。     

基于启发式调度规则的能力需求计划算法

传统的能力需求计划方法不能准确地反映实际生产能力需求与负荷.为此设计了基于启发式调度规则的能力需求计划算法,该算法将基于启发式规则的调度算法作用于物料需求计划进行能力需求计算,并对设备负荷率、关键设备资源和关键任务等关键能力信息的计算方法进行了重新定义,以更加真实地反映计划期内计划的能力需求与设备负荷状态.该算法已应用...     

一种动态识别瓶颈机床的前向启发算法

为了解决作业车间调度问题中的瓶颈漂移和动态识别问题,提出了一种动态识别瓶颈机床的前向启发算法.该算法引入搜索空间的概率分析方法,通过工序开工概率、工序对机床的独立需求概率、机床累计需求概率和3个量化参数,构造了以竞争高峰时段的机床负荷量大小为判定依据的瓶颈机床识别启发规则.实验结果表明该方法能够有效识别瓶颈机床,使得大规模作业车间调度的求解复杂度显著降低.     

在Linux下改进EDF实时调度，

EDF（Earliest Deadline First）算法由于CPU利用率高，可调度的任务集较大等优点在实时系统中的得到广泛的应用，但是EDF算法无法保证重要任务的执行。本文在EDF算法的基础上结合任务的重要性，提出了SBID（Scheduler Based Importance and Deadline）调度算法。首先从理论的角度分析了SBID算法的可调度性，并针对SBID算法对Linux2．6系统内核作一些修改。然后通过实验数据对比表明了SBID算法在保证重要任务顺利运行方面具有较好的优越性。     

基于维信息共享的粒子群优化算法在作业车间调度中的应用

为提高车间调度算法的寻优性能,提出了一种基于维信息共享的粒子群算法的车间调度问题解决方案。该算法对粒子群的认知过程和更新过程进行了研究,通过维信息共享和动态认知概念的引入,实现了优化问题维信息的沟通和交流,通过增加扰动因子克服算法的过早收敛,提高了对优化问题的适应能力。通过对3个连续函数优化问题的测试,得到了最佳的平均值和标准差,并对14个JSP标准测试案例进行仿真。结果表明无论是在求解质量还是收敛速度方面都优于其他几种算法,说明该算法能够有效地、高质量地解决作业车间调度问题。     

Task Scheduling Optimization in Cloud Computing by Rao Algorithm

Cloud computing is currently dominated within the space of high-performance distributed computing and it provides resource polling and on-demand services through the web. So, task scheduling problem becomes a very important analysis space within the field of a cloud computing environment as a result of user's services demand modification dynamically. The main purpose of task scheduling is to assign tasks to available processors to produce minimum schedule length without violating precedence restrictions. In heterogeneous multiprocessor systems, task assignments and schedules have a significant impact on system operation. Within the heuristic-based task scheduling algorithm, the different processes will lead to a different task execution time (makespan) on a heterogeneous computing system. Thus, a good scheduling algorithm should be able to set precedence efficiently for every subtask depending on the resources required to reduce (makespan). In this paper, we propose a new efficient task scheduling algorithm in cloud computing systems based on RAO algorithm to solve an important task and schedule a heterogeneous multiple processing problem. The basic idea of this process is to exploit the advantages of heuristic-based algorithms to reduce space search and time to get the best solution. We evaluate our algorithm's performance by applying it to three examples with a different number of tasks and processors. The experimental results show that the proposed approach significantly succeeded in finding the optimal solutions than others in terms of the time of task implementation.     

Impact of scheduling free goods on the throughput performance of a manufacturing operation

Theory of Constraints philosophy operates under the assumption that the goal of a for-profit business is to ‘Make money now and in the future’. Consequently, it is a manager's job continuously to search for ways to increase their operation's throughput. Increasing the amount of free goods the operation accepts is one way a manager can do this. Free goods are items that do not use the capacity of the operation's primary constraint. Because of this, it is possible to increase their output. However, increasing the number of free goods processed decreases the amount of protective capacity within the operation. Protective capacity plays a key role in the ability of an operation to deliver orders on time. Increasing the amount of free goods the operation processes is an action that could jeopardize the system's future throughput because it will impact on the operation's ability to ship orders on time. In order for managers to fulfil the goal of making money now and in the future, they must maximize total throughput while minimizing late orders. This paper provides some initial insights about how managers can intelligently manage free goods to balance these competing goals.     

多品种小批量订单型企业生产调度优化

目的研究多品种小批量订单型企业的生产调度优化问题,方法针对S公司的生产现状,应用遗传算法思想设计调度优化方案,采用不等长矩阵的编码方式实现订单的批量生产及车间排产的方案。结果通过仿真分析和S公司生产调度的实际应用,验证了该算法的可行性及有效性。结论基于遗传算法的调度优化算法实现了多品种小批量流程型生产企业生产调度优化,达到了缩短生产周期、有效利用生产资源的目的。     

A Joint Algorithm for Resource Allocation in D2D 5G Wireless Networks

With the rapid development of Internet technology, users have an increasing demand for data. The continuous popularization of traffic-intensive applications such as high-definition video, 3D visualization, and cloud computing has promoted the rapid evolution of the communications industry. In order to cope with the huge traffic demand of today’s users, 5G networks must be fast, flexible, reliable and sustainable. Based on these research backgrounds, the academic community has proposed D2D communication. The main feature of D2D communication is that it enables direct communication between devices, thereby effectively improve resource utilization and reduce the dependence on base stations, so it can effectively improve the throughput of multimedia data. One of the most considerable factor which affects the performance of D2D communication is the co-channel interference which results due to the multiplexing of multiple D2D user using the same channel resource of the cellular user. To solve this problem, this paper proposes a joint algorithm time scheduling and power control. The main idea is to effectively maximize the number of allocated resources in each scheduling period with satisfied quality of service requirements. The constraint problem is decomposed into time scheduling and power control subproblems. The power control subproblem has the characteristics of mixed-integer linear programming of NP-hard. Therefore, we proposed a gradual power control method. The time scheduling subproblem belongs to the NP-hard problem having convex-cordinality, therefore, we proposed a heuristic scheme to optimize resource allocation. Simulation results show that the proposed algorithm effectively improved the resource allocation and overcome the co-channel interference as compared with existing algorithms.     

求解一类无关并行机调度的遗传迭代贪心算法

以最小化最大完工时间为优化目标,建立带工单加工约束和序相关设置时间无关并行机调度问题的混合整数规划模型;考虑现实生产对求解算法在质量、收敛速度和鲁棒性等方面的较高要求,构建一种混合遗传-迭代贪心算法。在遗传变异操作中嵌入一种迭代贪心策略的破坏和构建机制,用于提高算法的种群多样性;引入基于破坏与构建操作设计而成的快速局部搜索算法来增强算法的局部开发能力;基于实际生产数据的相关特征随机生成了一系列计算案例,并通过实验说明所提新型混合算法相较于传统混合算法的优越性。     

Distributed supply‐chain project rescheduling: Part I—impacts of information‐sharing strategies

A supply‐chain project normally involves a number of independent and autonomous enterprises that share information to varying levels. The initial project schedule (preschedule) established at the time of forming the supply‐chain often requires a series of amendments due to unexpected or abrupt disturbances such as temporary resource outage (e.g. machine break), arrival or cancellation of orders from customers, and change in an operation's processing time (e.g. rework). Rescheduling or adaptive scheduling is a process of updating/repairing the preschedule to adapt to the disturbances. Appropriate rescheduling methods must be chosen and applied depending on the specific protocol of sharing information agreed between the enterprises in the supply chain. This paper is concerned with the impacts of different levels of information sharing on the performance of supply‐chain project rescheduling problems. Three rescheduling methods are examined in the research. They are distributed AOR (Affected Operations Rescheduling), negotiation‐based rescheduling (NEG), and centralized total rescheduling (TR). These three rescheduling algorithms represent three typical information‐sharing scenarios: little information sharing, limited information sharing and complete information sharing, respectively. A comprehensive computational study is conducted under different experimental settings. The results show that NEG and distAOR outperforms the TR rescheduling in terms of total cost minimization and stability of schedule and contractual relationship. NEG is superior in both rescheduling efficiency and effectiveness due to the effect of a moderate level of information sharing.     

Minimizing makespan subject to minimum total absolute deviation of completion time on identical parallel machines

This study addresses the identical parallel machine scheduling problem with the objective of minimizing makespan subject to minimum total absolute deviation of job completion time (TADC). An optimization algorithm is first proposed to solve TADC on an identical parallel machine and an iterative procedure based on a polynomial binary integer programming model is then proposed to minimize makespan. Computational experiments show that the proposed algorithm is efficient. The worst case performance, which refers to the largest average execution for each scenario of the experiments, is 229.10 seconds for the problem with n=200, m=30 and p _j from a uniform [1, 100].     

基于Kano模型的自理老人淋浴空间需求权重研究

目的针对产品设计初期用户需求度模糊的问题,提出了基于Kano模型的用户需求评价方法。旨在定性、定量分析自理老人淋浴空间需求,使用户需求与产品客观表现之间能够精确映射,更大程度地满足自理老人对淋浴空间安全、保健、易用等方面的深层次需求。方法借助Kano模型获取自理老人淋浴空间的需求信息,并对各需求项进行Kano类别的定性分类及筛选,构建自理老人淋浴空间需求层次结构。应用熵值法分别计算各级需求指标权重,进而获得各需求项的综合权重及重要度排序。结论对自理老人淋浴空间需求权重的研究,能准确分析用户需求偏好,为自理老人淋浴空间设计提供设计重点,提高自理老人对淋浴空间的满意度及用户体验,并为类似功能配置方案提供理论依据。     

A dynamic scheduling algorithm for singe-arm two-cluster tools with flexible processing times

This article presents a dynamic algorithm for job scheduling in two-cluster tools producing multi-type wafers with flexible processing times. Flexible processing times mean that the actual times for processing wafers should be within given time intervals. The objective of the work is to minimize the completion time of the newly inserted wafer. To deal with this issue, a two-cluster tool is decomposed into three reduced single-cluster tools (RCTs) in a series based on a decomposition approach proposed in this article. For each single-cluster tool, a dynamic scheduling algorithm based on temporal constraints is developed to schedule the newly inserted wafer. Three experiments have been carried out to test the dynamic scheduling algorithm proposed, comparing with the results the ‘earliest starting time’ heuristic (EST) adopted in previous literature. The results show that the dynamic algorithm proposed in this article is effective and practical.     

基于IC卡识别通勤客流的公交线路多模式调度

为解决现有的公交调度方式忽略了线路站点间空间客流分布的不均匀性,使得线路运行效率不高,本文提出了一种根据通勤客流构建单线路的多种调度模型,并设计相应算法,求解出在时间上和空间上与客流需求匹配调度模式。最后选取绵阳市一条实际的公交线路,进行案例分析,结果表明,通过该方式,高峰时段乘客总体时间比原来减少了5 290.9 min,优化效率高达13.93%,表明该方法能够节约乘客出行时间,提高线路运行效率。