超级计算机网络引导技术研究与分析

针对超级计算机系统中网络引导时间开销大的问题，提出网络引导分布算法是影响网络引导性能的主要因素之一，是优化网络引导性能的主要方向的观点。首先，分析了影响大规模网络引导性能的主要因素；其次，结合一种典型超级计算机系统，分析了超节点循环分布算法（SCDA）和插件循环分布算法（BCDA）的网络引导数据流拓扑结构；最后，量化分析了这两种算法对各个网络路径段的压力和可获得的网络性能，发现BCDA性能是SCDA性能的1~20倍。通过理论分析和模型推导发现，在计算节点和引导服务器之间使用更细粒度的映射算法可以在引导部分资源时使用尽量多的引导服务器，减少对局部网络资源的过早竞争，提升网络引导性能。     

Modeling and maximum power point tracking (MPPT) method for PV array under partial shade conditions

Influenced by partial shade, PV module aging or fault, there are multiple peaks on PV array's output power–voltage (P–V) characteristic curve. Conventional maximum power point tracking (MPPT) methods are effective for single peak P–V characteristic under uniform solar irradiation, but they may fail in global MPP tracking under multi-peak P–V characteristics. Existing methods in literature for this problem are still unsatisfactory in terms of effectiveness, complexity and speed. In this paper, we first analyze the mathematical model of PV array that is suitable for simulation of complex partial shade situation. Then an adaptive MPPT (AMPPT) method is proposed, which can find real global maximum power point (MPP) for different partial shade conditions. When output characteristic of PV array varies, AMPPT will adjust tracking strategies to search for global peak area (GPA). Then it is easy for conventional MPPT to track the global MPP in GPA. Simulation and experimental results verify that the proposed AMPPT method is able to find real global MPP accurately, quickly and smoothly for complex multi-peak P–V characteristics. Comparison analysis results demonstrate that AMPPT is more effective for most shade types.     

Maximum power point tracker based on maximum power point resistance modeling

This paper presents a new, simple, accurate, and inexpensive practical methodology and experimental solution for the modeling of conventional (domestic and commercial facilities) photovoltaic generators (PVG), so that they can work at their maximum power point (MPP). The PVG may be a panel, an array of panels, or a photovoltaic field. As a starting restriction (actual) it has to be assumed that it is not possible to isolate the PVG variables dependence (mainly current, voltage, and hence power) with solar radiation and temperature, because they are highly correlated. This methodology proposes the modeling facility on its MPP by its MPP resistance (R_MPP = V_MPP/I_MPP), being V_MPP and I_MPP the voltage and current of the PVG, respectively, at its MPP. The analysis shows, by simulation first and then experimentally, that R_MPP does not present significant temperature dependencies, at least in the usual range. This important result allows us to model the PVG only in terms of solar radiation. From a set of experimental data, different models to estimate R_MPP are proposed. The obtained results are very accurate. These models allow an immediate practical application that it is also developed in the paper: MPP tracker (MPPT) design by the calculation of the DC/DC converter duty cycle which places the PVG at its MPP directly and continuously. This new methodology and experimental system has been registered in the Spanish Patent and Trademark Office with the number P201530352. Copyright © 2015 John Wiley & Sons, Ltd.     

Toward multiple maximum power point estimation of photovoltaic systems based on semiconductor theory

Environmental conditions, such as temperature, non‐uniform irradiation, and solar shading, deeply affect the characteristics of photovoltaic (PV) modules in PV‐assisted generation systems. Several local maximum power points (MPPs) are found in the power–voltage curve of PV systems constructed by series/parallel‐connected PV modules under partially shaded conditions. The characteristics of PV systems change unpredictably when multiple MPPs occur, so the actual MPP tracking (MPPT) becomes a difficult task. Conventional MPPT methods for the PV systems under partially shaded conditions cannot quickly find the actual MPP such that the optimal utilization of PV systems cannot be achieved. Based on the p–n junction semiconductor theory, we develop a multipoint direct‐estimation (MPDE) method to directly estimate the multiple MPPs of the PV systems under partially shaded conditions and to cope with the mentioned difficulties. Using the proposed MPDE method, the multiple MPPs of the PV systems under partially shaded conditions can be directly determined from their irradiated current–voltage and power–voltage characteristic curves. The performances of the proposed MPDE method are evaluated by examining the characteristics of multiple MPPs of PV systems with respect to different shading strengths and numbers of the shaded PV modules and also tested using the field data. The experimental results demonstrate that the proposed MPDE method can simply and accurately estimate the multiple MPPs of the PV systems under partially shaded conditions. The optimization of MPP control models and the MPPT for PV systems could be achieved promisingly by applying the proposed method. Copyright © 2014 John Wiley & Sons, Ltd.     

Structurally laminated CrN films deposited by multi pulse modulated pulsed power magnetron sputtering

The paper presents the results on the deposition of nanoscale structurally laminated CrN films using a novel multi pulse modulated pulsed power (MPP) magnetron sputtering technique. With the multi pulse MPP approach, thin films with a structural modulation in the nanometer range are obtained by alternately switching two (or even more) high power MPP pulses on the same target, which have different pulse lengths, frequencies and powers. Each pulse was turned on for a pulse repeat duration during which this given pulse shape was repeated. In this study, CrN films have been deposited in a closed field unbalanced magnetron sputtering system using the multi pulse MPP technique by varying the pulse repeat duration of two different pulses. The CrN films were also deposited by dc magnetron sputtering (dcMS) and single pulse MPP techniques for comparison. The microstructure and properties of the films were characterized using glancing incident X-ray diffraction, scanning electron microscopy, transmission electron microscopy, nanoindentation, and ball-on-disk wear tests. The structure and properties of the multi pulse MPP CrN films depended on the pulse repeat duration. The highest hardness of 30.5 GPa and an H/E ratio of 0.9 have been achieved in the multi pulse MPP CrN films. The wear rate of the single pulse MPP and multi pulse MPP CrN films decreased by a factor of 5.8–17 as compared to the dcMS CrN films.     

Two-dimensional nonsteady viscous flow simulation on the Navier-Stokes Computer miniNode

The needs of large-scale scientific computation are outpacing the growth in performance of mainframe supercomputers. In particular, problems in fluid mechanics involving complex flow simulations require far more speed and capacity than that provided by current and proposed Class VI supercomputers. To address this concern, the Navier-Stokes Computer (NSC) was developed. The NSC is a parallel-processing machine, comprised of individual Nodes, each comparable in performance to current supercomputers. The global architecture is that of a hypercube, and a 128-Node NSC has been designed. New architectural features, such as a reconfigurable many-function ALU pipeline and a multifunction memory-ALU switch, have provided the capability to efficiently implement a wide range of algorithms. Efficient algorithms typically involve numerically intensive tasks, which often include conditional operations. These operations may be efficiently implemented on the NSC without, in general, sacrificing vector-processing speed. To illustrate the architecture, programming, and several of the capabilities of the NSC, the simulation of two-dimensional, nonsteady viscous flows on a prototype Node, called the miniNode, is presented.     

LS MPP编程语言研究

为了方便开发运行于LS MPP系统的应用程序,文章研究了LS MPP编程语言.首先,分析了现有LS MPP计算机的体系结构,以及作为其未来发展方向的图像处理器的概念模型.然后,介绍了对应该概念模型的中间语言和中间表示.最后,详细分析了该概念模型对应的高级语言扩展部分.分析表明,高级语言对概念模型描述的并行计算机性能的提高非常有益,使程序员的编程更加方便,并且降低了编译器的复杂性.     

Design and optimization of parallel computers for processing mass data

Problems of high-performance processing of mass cluster data are considered. Estimates of execution times of parallel data processing programs and a heuristic algorithm of optimization of cluster architectures for such problems are proposed. __________ Translated from Kibernetika i Sistemnyi Analiz, No. 4, pp. 117–129, July–August 2006.     

PVM与网络并行计算

松散耦合的异构型并行处理系统近年来又有较大的发展，这种只能支持粗粒度并行的计算环境，由于通讯硬件的更新和软件环境的开发，其性能已经能够与某些ＭＰＰ机相媲美，因而形成并行处理领域一个强有力的分支－网络计算，本文围绕ＰＶＭ，讨论网络计算的软件环境，概述其特色和发展，并与其它基于消息传递的软件环境相比较，最后预测ＰＶＭ未来可能面临的问题和发展方向。