光学可重排全交叉—逆全交叉网络及其与Benes网络拓扑等价的…

互连网络是数字光计算机及并行电子计算机体系结构的重要研究课题。本文提出了可重排全交叉－逆全交叉网络的拓扑结构及其光学实现方案。并采用互连网络拓扑等价的图分析法得到了全交叉－逆全交叉网络与Ｂｅｎｅｓ网络具有拓扑等价性质的多套逻辑名结构。为开拓光学可重排人交叉－逆全交叉网络在光电混合巨型并行多处理计算机系统潜在应用提供了理论依据。     

光学可重排全交叉—逆全交叉网络及其与Benes网络拓扑等价的多套逻辑名结构

互连网络是数字光计算机及并行电子计算机体系结构的重要研究课题。本文提出了可重排全交叉—逆全交叉网络的拓扑结构及其光学实现方案。并采用互连网络拓扑等价的图分析法得到了全交叉—逆全交叉网络与Benes网络具有拓扑等价性质的多套逻辑名结构。为开拓光学可重排全交叉—逆全交叉网络在光电混合巨型并行多处理计算机系统等领域的潜在应用提供了理论依据。     

一种用于机群系统的双环光互连网络性能分析

高性能计算机网络对带宽的需求,使得如何提供高带宽的互连网络以及充分利用互连网络的固有带宽度为一个研究焦点。基于Linux操作系统,以千兆光互连接口卡为网络接口,设计实现了一种可用于机群系统的双环光互连网络。介绍了光互连网接口卡的基本结构,接口卡驱动软件的设计方法,双环光互连网络的拓朴结构及其特性。分析测试了光互连网络的通信性能,指出了影响系统整体性能的关键因素。     

SuperStar:一种可扩展高阶互连拓扑结构

随着高速信号传输技术和VLSI技术的发展,使用高阶路由器来应对因高性能计算机峰值性能不断攀升给高性能互连网络带来的新挑战已是发展需求;同时,如何利用高阶特性减少互连网络延迟和成本开销,以支持更大的网络规模是设计高性能互连网络拓扑结构的关键和突破点。针对目前基于高阶路由器的典型拓扑结构进行了分析,并在此基础上提出一个新的高阶拓扑架构SuperStar,其不仅具有较短的网络直径而且具有良好的可扩展性;通过在一个基于OMNeT++平台自主开发的高阶互连网络性能测评模拟器上设定不同的通信负载,测评各种拓扑结构在通信系统下实际的网络延迟和吞吐率的走势,以分析SuperStar的通信开销。     

VTFTR：高维胖树中的无死锁容错路由算法

随着近年来高性能计算系统规模的急剧扩大,高性能互连网络的可靠性成为愈发重要的问题。高维胖树是一种结合了胖树与多维环网优点的网络拓扑结构,凭借其良好的可扩展性与网络性能在E级时代具有广阔的应用前景。然而,目前关于高维胖树中容错路由算法的相关研究较为有限,其可靠性问题亟待解决。为提高高维胖树拓扑在高性能互连网络中的容错能力,进一步提高对应超算系统的运行效率,提出一种用于高维胖树中叶交换机故障的容错路由算法VTFTR。该算法结合转向模型与虚通道切换的思想,通过严格控制报文在无故障路径与容错路径中的转向,使用少量的容错虚通道与额外跳步实现高维胖树中的无死锁容错。实验结果表明,在单点故障情况下,VTFTR算法的容错路径较对比算法有2~4个跳步的减少,在4 096个节点规模的网络中,当叶交换机故障数量为10时,在故障叶交换机不同的分布情况下,该算法能够以1.4%~2.0%的吞吐率下降作为代价来保持全网无故障节点之间的互连。     

面向HPC和DC的可重构光互连网络体系结构综述

互连网络是高性能计算系统和数据中心的核心组件之一,也是决定其系统整体性能的全局性基础设施。随着高性能计算、云计算和大数据技术的迅速发展,传统的电互连网络在性能、能耗和成本等方面无法满足高性能计算应用和数据中心业务的大规模可扩展通信需求,面临着严峻的挑战。为此,近年来相关研究者提出了多种面向高性能计算和数据中心的可重构的光互连网络结构。首先阐明了光互连网络相对于电互连网络的优势;然后介绍了几种典型的可重构光互连网络体系结构,并对其特点进行了分析比较;最后探讨了可重构光互连网络的发展趋势。     

双环网嵌入RP(k)网络

网络嵌入是互连网络研究的一个重要方向,通过网络嵌入可以用一种拓扑结构模拟另一种结构,高效的嵌入会提高并行程序的运行效率。构造了10*k个节点的双环网结构,基于文献[3]提出的互连网络RP（k）,提出了一种将双环网嵌入RP（k）的算法DLN-RP（k）,此算法得到的4个性能参数为拓展、负载、延伸、拥挤度分别为1,1,2,2,并证明了该结果为最优值。     

新型互连网络NIN研究

以通信延迟和网络吞吐率这两个重要参数为出发点,以大规模并行处理系统的新型互连网络（ＮＩＮ）进行了理论分析和模拟测试,并与二维网格、环形网以及反图拓扑互连网络进行了比较,结果表明：与其它三种网络相比,ＮＩＮ不仅保持了反图拓扑互连网络可以连接更多处理机的优点,同时还具有较短的通信延迟和较高的网络吞吐率。     

典型TCP/IP网络互连的排错分析及测试过程

连接不同的网络已经成为网络工程中必不可少的基础性环节,由于网络互连需要的设备数量众多,需要配置大量的参数,给网络测试和网络排错带来了一定的复杂度。本文以一个典型TCP/IP网络互连拓扑为例,进行了典型网络互连的排错分析,并设计了规范的测试过程。     

超立方体网络的不相交路径通信策略研究综述

超立方体是一类具有良好的拓扑性质的互连网络模型。不相交路径的实现是超立方体网络中容错通信的有效保证。介绍了超立方体网络的不相交路径路由策略中的主要研究内容和研究现状,对近年来该方面取得的研究成果进行分析和总结,并指出了其中存在的问题和该策略研究的方向。     

超立方体双环互连网络及路由算法*

给出了一种可扩展的互连网络拓扑结构,称为超立方体双环。该互连网络拓扑结构结合了超立方体拓扑的短直径、高连通性、对称性、路由简单和一种新的双环拓扑结构的可扩展性和常数节点度的优点,使得网络规模增大时,网络节点度可以保持常数;网络节点采用格雷编码和约翰逊编码的混合编码方法,网络的任意相邻节点编码有且仅有一位不同,使得路由算法设计简单。最后分别设计了基于混合编码的单播、广播路由算法。分析表明提出的互连网络具有较好的拓扑性质和通信性能。     

The OTIS hyper hexa-cell optoelectronic architecture

Optical transpose interconnection system (OTIS) is an optoelectronic architecture that promises to be a great choice for future-generation parallel systems. OTIS combines the advantages of electronic and optical links, where electronic links are used for short distances which require low material cost, and optical links are used for long distances which provide high speed network with low power consumption. Taking into account the advantageous characteristics of OTIS and based on the attractive properties of hyper hexa-cell (HHC) interconnection topology from low diameter and good minimum node degree, this paper introduces a new optoelectronic architecture referred to as OTIS hyper hexa-cell (OHHC). This paper also provides an evaluation and a comparison of the new topology with OTIS-mesh in terms of the following topological properties: size, diameter, maximum and minimum node degree, bisection width, total cost and optical cost. The results of this study proved the excellence of the proposed OHHC over OTIS-mesh in terms of diameter, minimum node degree, bisection width, and optical cost.     

并行LU分解的通信模式在WDM环网上的波长分配算法

波长分配是光网络设计的基本问题，设计波长分配算法是洞察光网络通信能力的基本方法．不同的并行算法具有不同的通信模式，如何在光互连网上实现这些通信模式，是当前一个颇受关注的研究领域．本文基于WDM环网络，针对矩阵的并行LU分解，构造了一种并行LU分解的通信模式，讨论了将该通信模式嵌入在环形光网络中的波长分配问题．在解决该问题的过程中，得到了将一种特殊的二分图结构的通信模式嵌入在环网中的波长分配算法．通过分析和证明得到了在WDM环网上实现该并行LU分解通信模式所需的最小波长数．     

无缓冲光互连网络的延时性能分析及优化

针对当前高性能计算机光互连网络中光缓冲不易实现的问题,提出一种无缓冲的光互连网络结构BOIN,在对网络结构进行建模和分析的基础上,研究网络的传输延时随不同输入负载和网络规模而变化的规律,给出在一定互连总规模和输入负载下,网络延时达到最小值时网络拓扑结构所必须满足的条件。模拟实验证明了该结果的正确性。     

基于InfiniBand的多链路mesh/torus大规模并行系统互连网络

在大规模并行系统中,系统级互连网络的设计至关重要.InfiniBand作为一种高性能交换式网络被广泛应用于大规模并行处理系统中.mesh/torus拓扑结构相较于目前普遍应用于InfiniBand网络的胖树拓扑结构拥有更好的性能与可扩展性.尽管如此,研究发现,用传统的mesh/torus拓扑结构构建InfiniBand互连网络存在诸多问题.分析了传统网络拓扑结构的缺陷,并提出了一种基于InfiniBand的多链路mesh/torus互连网络.这种改进型的拓扑结构通过充分利用交换机间的多链路可以获得比传统mesh/torus网络更高的带宽.另外,同时给出了与该网络拓扑结构相配套的高效路由算法.最后,通过网络仿真技术对提出的算法进行了评估,实验结果显示提出的路由算法相较于其他路由算法拥有更好的性能与可扩展性.     

A scatternet formation algorithm for Bluetooth networks with a non-uniform distribution of devices

We address the problem of scatternet formation for network scenarios with a non-uniform distribution of Bluetooth devices. The assumption of having a non-uniform distribution of devices in a given area of interest is motivated by examples that can be encountered in real scenarios. We propose a new scatternet formation protocol called BlueHRT (Bluetooth Hybrid Ring Tree) that results in a hybrid ring tree topology. The proposed protocol operates in multiple phases that include node discovery, identification of a dense area within the network, role assignment with ring based piconets in the dense area and tree based piconets in the surrounding lightly loaded areas, and interconnection of the piconets to each other via slave–slave and master–slave bridges. General protocol design analysis and ns-2 simulation results are presented in order to highlight the main performance characteristics.     

A spanning multichannel linked hypercube: a gradually scalableoptical interconnection network for massively parallel computing

A new, scalable interconnection topology called the Spanning Multichannel Linked Hypercube (SMLH) is proposed. This proposed network is very suitable to massively parallel systems and is highly amenable to optical implementation. The SMLH uses the hypercube topology as a basic building block and connects such building blocks using two-dimensional multichannel links (similar to spanning buses). In doing so, the SMLH combines positive features of both the hypercube (small diameter, high connectivity, symmetry, simple routing, and fault tolerance) and the spanning bus hypercube (SBH) (constant node degree, scalability, and ease of physical implementation), while at the same time circumventing their disadvantages. The SMLH topology supports many communication patterns found in different classes of computation, such as bus-based, mesh-based, and tree-based problems, as well as hypercube-based problems. A very attractive feature of the SMLH network is its ability to support a large number of processors with the possibility of maintaining a constant degree and a constant diameter. Other positive features include symmetry, incremental scalability, and fault tolerance. It is shown that the SMLH network provides better average message distance, average traffic density, and queuing delay than many similar networks, including the binary hypercube, the SBH, etc. Additionally, the SMLH has comparable performance to other high-performance hypercubic networks, including the Generalized Hypercube and the Hypermesh. An optical implementation methodology is proposed for SMLH. The implementation methodology combines both the advantages of free space optics with those of wavelength division multiplexing techniques. A detailed analysis of the feasibility of the proposed network is also presented     

光RP(k)网络上Hypercube通信模式的波长指派算法

波长指派是光网络设计的基本问题,设计波长指派算法是洞察光网络通信能力的基本方法.基于光RP(k)网络,讨论了其波长指派问题. 含有N=2ⁿ个节点的Hypercube通信模式,构造了节点间的一种排列次序Xn,并设计了RP(k)网络上的波长指派算法.在构造该算法的过程中,得到了在环网络上实现n维Hypercube通信模式的波长指派算法.这两个算法具有较高的嵌入效率.在RP(k)网络上,实现Hypercube通信模式需要max{2,「5(2^n-5/3」}个波长.而在环网络上,实现该通信模式需要复用(N/3+N/12(个波长,比已有算法需要复用「N/3+N/4」个波长有较大的改进.这两个算法对于光网络的设计具有较大的指导价值.     

机群系统DDR DIMM总线光互联网络适配器设计

介绍了一种面向机群系统双环形网络拓扑结构的高速光互联网络适配器的设计和实现方法。该网络适配器基于FPGA技术实现，总线接口采用高速、高带宽的DDR DIMM总线，网络传输介质采用光纤，底层路由协议采用FPGA内部硬件逻辑实现，全方位保证了高带宽、低延迟、高可靠的网络特性。     

The Optical Chained-Cubic Tree interconnection network: Topological structure and properties

Interconnection networks with optical communication links outperform others using electronic communication links when the distance is long in terms of speed and power consumption. However, for short distances, electronic network topologies are preferred due to lower material cost requirements. As a result, hybrid network topologies were constructed to combine the benefits of both types of network topologies, such as Optical Transpose Interconnection System (OTIS). This paper presents a new hybrid interconnection network topology, which is constructed using both optical and electronic links, called the Optical Chained-Cubic Tree (OCCT). This new OCCT topology is based on the Chained-Cubic Tree (CCT) interconnection network and is designed to cope with both types of binary trees; full and complete. Also, the topological properties of OCCT in terms of diameter, connectivity, degree, bisection width, and cost are presented and compared with OTIS-Mesh and CCT interconnection networks.