费用-时间优化的网格有向无环图调度算法

网格环境下,基于时间限制和费用最小的有向无环图(DAG)调度算法运用经济规律把网格用户中的任务映射到网格资源中运行,弥补了Buyya R提出的算法中未考虑任务运行的优先关系.该算法有两个关键技术:DAG中有效路径的提取能够定位任务何时在资源上运行;DAG中在作业运行时间尽可能允许的情况下,把任务映射到价格便宜的资源上运行.通过仿真实例,论证了该算法的优越性.     

网格调度的算法研究及改进

优化网格资源调度算法可以提高网格系统执行效率,给任务安排合理的执行顺序和合适的处理器是优化网格资源调度算法需突破的关键技术.文中研究并实现了（Heterogeneous-Earliest-Finish） HEFT[1]算法和新的（Hierarchical Reliability-Driven Scheduling）HRDS算法.采用DAG[2]任务图生成函数,通过对已有HEFT算法进行研究,采用SimGrid为在分布计算环境下进行分布并行应用调度研究提供一个仿真环境,对HRDS算法进行了改进和验证.验证过程中在HRDS算法中加入了可靠性开销作为调度依据,并把算法分为两层调度,本地可靠性驱动调度和全局可靠性驱动调度.两算法的调度结果在SimGrid网格模拟器中仿真调度,仿真成功并且调度结果在可靠性和性能方面HRDS都比HEFT算法要好.     

树形异构网格的启发任务调度算法

讨论了在网格资源计算能力和网络通信速度异构的树形网格环境下任务调度问题,导出了线性方程并且根据调度任务大小进行了模型的优化,提出一个基于线性规划的任务分配启发式算法。实验结果表明:在异构树形计算网格环境下实现任务调度时,该算法的性能明显优于其他算法。     

包含多重嵌套轮廓线的激光切割空移路径规划

包含多重嵌套轮廓线的空移路径规划是开发激光切割系统的主要问题之一,在满足嵌套图形由内到外的激光切割工艺要求下,提出启发式排序和网格排序算法,实现优化排序。首先通过射线法,判断多重嵌套轮廓线的位置关系;然后采用最佳适应度优先的启发式排序算法,将空移距离、轮廓线的面积作为评价指标,分别赋值权重得到总体适应度,选择适应度高的轮廓线作为下一个切割图形;最后,为了满足不同应用场景,提出另外一种网格排序方法,根据轮廓线控制点的疏密程度,基于层次聚类算法划分网格,按照规则的路线遍历网格,依次确定定位到每个网格中的轮廓线。仿真与试验结果表明,相比于智能优化排序算法,启发式排序和网格排序在满足激光切割工艺要求的前提下,不仅可以有效缩短空移路径,还大大减少了计算时间,显著提高激光切割效率和质量。     

波分复用光网中的一种新型波长分配算法

目前网络承载业务的趋势是以IP为中心的数据业务,从而促进了以WDM光网络向高速和宽带多波长的应用和普及,为了进一步提高网络的性能并提高资源利用率,出现了光网络控制面技术.路由选择与波长分配问题是WDM光传输网络控制面中非常重要的问题之一.目前实际应用最广泛的波长分配算法是First-Fir(FF)算法.本文基于FF算法,研究动态业务下波分复用光网络在固定选路下的波长分配问题,提出了一种新的波长分配启发式算法——Joint First Fit.计算机仿真试验表明,与常用的FF算法相比,该算法显著的降低了网络呼叫阻塞率,有利于提高网络资源的利用率.     

网格计算中一种启发式数据分配算法的讨论

网格是在高速网络环境下解决大区域、高性能、并行计算问题的一种新技术.高性能调度技术是网格核心管理的重要技术之一,它要做的第一步就是在空间上对计算和数据进行分配.本文针对一个简化了的网格计算环境,考虑待处理的数据单元在各聚集中出现的不同频率,提出了一种启发式数据分配算法(HDDA),可以使数据分配在多项式时间内完成并获取较小的通信费用,然后对算法的进行了性能分析,最后对该算法在实际网格中的应用进行了探讨.     

基于生存性和Makespan的多目标网格任务调度算法研究

在动态、复杂的网格系统中,资源的失效非常频繁,网格资源的失效会导致在该节点上执行的计算任务无法正常完成,从而影响网格计算的服务质量和效率。针对这个问题,提出了任务生存性(survivability)的概念,将任务生存性与Makespan结合起来,给出了一个可调节的局部目标函数,实现了基于生存性和Makespan的Min-min启发式调度算法(SM-Min-min)。实验证明该算法能够有效地平衡Makespan和任务生存性目标,并可以通过调节目标函数中的偏好参数满足不同的需求,因此更适合开放复杂的计算网格环境。     

基于连接强度的PPI网络蚁群优化聚类算法

 由于PPI网络数据的无尺度和小世界特性,使得目前对此类数据的聚类算法效果不理想.根据PPI网络的拓扑结构特性,本文提出了一种基于连接强度的蚁群优化(Joint Strength based Ant Colony Optimization,JSACO)聚类算法,该算法引入了连接强度的概念对蚁群聚类算法中的拾起/放下规则加以改进,以连接强度作为拾起规则,对结点进行聚类,并根据放下规则放弃部分不良数据,产生最终聚类结果.最后采用了MIPS数据库中的PPI数据进行实验,将JSACO算法与PPI网络数据的其他聚类算法进行比较,聚类结果表明JSACO算法正确率高,时间开销低.     

基于混合并行遗传算法的网格资源分配策略

网格是利用互联网或专用网络将地理上广泛分布的、异构的、动态的资源互联起来实现资源高度共享与集成，为用户提供高性能的计算、管理和服务等功能的一项新技术。但如何将网格这个复杂环境中的资源有效进行管理和调度，是一个NP难问题，同时也是网格技术发展的关键。启发式算法被证明是解决这类问题的有效算法，将两个启发式算法结合起来，充分发挥各自的优势，就能有效解决网格资源分配的问题。因此，利用混合并行遗传算法来解决网格资源分配是可行的。     

概率数据流上Skyline查询处理算法

 概率数据流管理与分析逐步引起了研究者们的关注.Skyline查询技术是近年来数据库领域的研究热点.此前相关工作仅限于静态数据集或传统确定性数据流上的Skyline查询处理,尚无人考虑概率数据流上的Skyline计算问题,本文提出的SOPDS算法则较好地解决了该问题.在采用适应性更强的网格索引的基础上,提出了概率定界、逐步求精、提前淘汰与选择补偿等启发式规则对算法从时间和空间两方面进行了系统地优化.实验表明,算法在时间与空间上具有较高的整体性能.     

Approach of Balancing of the Negotiation among Agents in Traffic Synchronization

This paper presents a new solution for Air Traffic Flow Management (ATFM) problem. The main objective of our research is to improve computational efficiency for ATFM by combing grid computing with multi-agent coordination techniques. To balance the communication among Multi-Agents in ATFM for grid computing environments, ATFM in Grid Computing (ATFMGC) is proposed for such synchronization problem. To demonstrate the developed model, the function of the agents, their knowledge representation and inference processes are discussed and the grid architecture in which the agents execute is described. Also, as a criterion to measure the effectiveness to reduce the amount of communication among agents and the delay of the flights, a metric called Standard of Balancing among Agents (SBA) is used as a basic index. The simulation results show the efficiency of the developed model and its successful application in a case study.     

协方差矩阵稀疏表示在网格失配波达方向估计中的应用

该文利用了入射信号在空域的稀疏性,将波达方向（DOA）估计问题描述为在网格划分的空间协方差矩阵稀疏表示模型,并将其松弛为一个凸问题,从而提出了一种网格匹配下的交替迭代方法（AIEGM）。传统的基于稀疏重构的波达方向估计算法由于其模型的局限性,一旦入射角不在预先设定的离散化网格上,就会造成估计性能的急剧恶化。针对这个问题,该算法可以在离散化网格比较粗糙的前提下,通过交替迭代的方法求解一系列基追踪去噪（BPDN）问题,对于不在网格上的真实角度估计值进行修正,从而达到更精确的波达方向估计。仿真结果证明了AIEGM算法的有效性。      

网格工作流动态调度研究

现有的网格工作流调度算法大都利用遗传算法所具有的并行性和全局解空间搜索的特点来解决工作流调度问题.但是,现有的调度算法没有对动态调度问题进行处理.文中针对网格服务的动态性,提出了服务资源信息中心的概念并给出了网格工作流管理系统的体系结构.在现有的基于遗传算法的网格工作流调度算法的基础上提出了网格服务工作流动态调度算法,补充了不同工作流过程模型的适应度函数的计算.     

On the canonical grid method for two-dimensional scatteringproblems

The banded matrix iterative approach with a canonical grid expansion (BMIA/CAG) has been shown by Tsang et al. (see ibid., vol.43, p.851-9, 1995) to be an efficient method for the calculation of scattering for near planar two-dimensional (2-D) geometries such as one-dimensional rough surfaces. However, in the article of Tsang et. al, only the first three terms in the canonical grid series for TE polarization above a perfectly conducting surface were discussed and methods for implementing only a portion of these terms were presented. In this paper, a general form for all terms in the canonical grid series is provided for both TE and TM polarizations above an impedance surface and an efficient algorithm for calculating their contributions is described. The relationship between the canonical grid and operator expansion methods is also discussed. A sample surface scattering problem is shown to illustrate the utility of higher order terms in the canonical grid method     

基于扰动的结合Off-grid目标的层析SAR三维成像方法

层析合成孔径雷达(TomoSAR)通过组合在不同高度上获取的多基线二维SAR数据,实现合成孔径雷达的三维成像.TomoSAR的求解本质是一维谱估计问题,基于压缩感知的方法可以在非均匀分布的少量基线观测下实现求解,逐渐成为了主流的成像方式.在经典的压缩感知算法流程中,需要将连续的高程向划分成固定的网格,并且假定目标正好位...     

Monte Carlo simulations of wave scattering from lossy dielectricrandom rough surfaces using the physics-based two-grid method and thecanonical-grid method

In using the method of moments to solve scattering by lossy dielectric surfaces, usually a single dense grid (SDG) with 30 points per wavelength is required for accurate results. A single coarse grid (SCG) of ten points per wavelength does not give accurate results. However, the central processing unit (CPU) and memory requirements of SDG are much larger than that of SCG. In a physics-based two-grid method (PBTG) two grids are used: a dense grid and a coarse grid. The method is based on the two observations: (1) Green's function of the lossy dielectric is attenuative and (2) the free-space Green's function is slowly varying on the dense grid. In this paper, the PBTG method is combined with the banded-matrix iterative approach/canonical grid method to solve rough surface scattering problem for both TE and TM cases and also for near grazing incidence. We studied cases of dielectric permittivities as high as (25+i)ϵ₀ and incidence angle up to 85°. Salient features of the numerical results are: (1) an SCG has poorer accuracy for TM case than TE case; (2) PBTG-banded-matrix iterative approach/canonical grid BMIA/CAG method speeds up CPU and preserves the accuracy; it has an accuracy comparable to single dense grid and yet has CPU comparable to single coarse grid; (3) PBTG-BMIA/CAG gives accurate results for emissivity calculations and also for low grazing backscattering problems (LGBA); and (4) the computational complexity and the memory requirements of the present algorithm are O(N log(N)) and O(N), respectively, where N is the number of surface unknowns on the coarse grid     

Active-fault-alarm enabled pre-protection scheme in 10G-EPON for smart power grid

The active-fault-alarm(AFA) has been developed to achieve fast recovery of various network services,but the false-alarm still exists when applied to 10 Gigabit ethernet positive optical network(10G-EPON) in power grid.To solve this problem,an active fault alarm based pre-protection(AFA-PP) scheme in 10G-EPON system for power grid is proposed in this paper.The AFA-PP scheme takes full advantage of the AFA technologies,and allows conducting pre-protection operation in advance.Analytical and experimental results suggest the better performances of the new mechanism and show that the AFA-PP can solve the false-alarm problem efficiently.     

Experimental wire grid modelling for antennas on structures

Theoretical wire grid modelling for antennas on structures involves the approximate solution of an approximate problem. By experimentally modelling the approximate problem of a wire grid replacing a continuous metal structure, it is shown that a wire grid with spacing ? ?/10 is a good approximation both for the far-field radiation pattern of the antenna and for its input impedance. Difficulties in providing accurate impedance predictions should therefore be ascribed to the approximate mathematical solution, rather than the wire grid model.     

RDMSR: A reactive defragmentation with minimum spectrum route strategy for mixed grid optical network

In this paper, we propose a reactive defragmentation with minimum spectrum route (RDMSR) for the problem of route, spectrum, and modulation-format allocation (RSMA) in mixed grid optical network. In mixed grid network, spectrum redundancy and its assignment problem increase due to the spectrum continuity and contiguity constrains. In the proposed RDMSR strategy, process of defragmentation initiates after the termination of the existing connections. In addition, the route that needs minimum spectrum is given priority over the other available routes. The performance of the proposed strategy is compared with the two existing strategies: shortest path first (SPF) and mixed grid aware dynamic resources allocation (MDRA). Simulations are performed on NSFNET and ARPANET topologies. The existing and proposed strategies are evaluated on the metrics of bandwidth blocking probability (BBP), network capacity (NC), and average hops (AH) at three different combinations of the network traffic. Results show the proposed strategy outperforms than the other existing state of art strategies.     

基于原子范数最小化的步进频率 ISAR一维高分辨距离成像方法

针对传统离散化压缩感知方法在网格失配条件下步进频率(SF) ISAR 1维距离成像估计性能下降的问题,该文提出一种基于原子范数最小化(ANM)的高分辨距离成像方法。首先,构建基于原子范数的无网格SF ISAR距离向稀疏表示模型,将1维距离成像问题转化为原子系数以及频率估计问题。然后,利用原子范数半正定性质,将原子范数最小化问题转化为半正定规划问题,并基于交替方向乘子法实现快速求解。最后,利用Vandermonde分解得到最终的1维高分辨距离成像结果。由于避免了网格离散化处理,因此可以实现网格失配、低量测值条件下的高分辨距离成像,且保持了高的距离分辨能力。理论分析与仿真实验验证了所提方法的有效性。