Scalable optical hypercube-based interconnection network for massively parallel computing

Two important parameters of a network for massively parallel computers are scalability and modularity. Scalability has two aspects: size and time (or generation). Size scalability refers to the property that the size of the network can be increased with nominal effect on the existing configuration. Also, the increase in size is expected to result in a linear increase in performance. Time scalability implies that the communication capabilities of a network should be large enough to support the evolution of processing elements through generations. A modular network enables the construction of a large network out of many smaller ones. The lack of these two important parameters has limited the use of certain types of interconnection networks in the area of massively parallel computers. We present a new modular optical interconnection network, called an optical multimesh hypercube (OMMH), which is both size and time scalable. The OMMH combines positive features of both the hypercube (small diameter, high connectivity, symmetry, simple routing, and fault tolerance) and the torus (constant node degree and size scalability) networks. Also presented is a three-dimensional optical implementation of the OMMH network. A basic building block of the OMMH network is a hypercube module that is constructed with free-space optics to provide compact and high-density localized hypercube connections. The OMMH network is then constructed by the connection of such basic building blocks with multiwavelength optical fibers to realize torus connections. The proposed implementation methodology is intended to exploit the advantages of both space-invariant free-space and multiwavelength fiber-based optical interconnect technologies. The analysis of the proposed implementation shows that such a network is optically feasible in terms of the physical size and the optical power budget.     

Incrementally scalable optical interconnection network with a constant degree and constant diameter for parallel computing

A new scalable interconnection topology called the spanning-bus connected hypercube (SBCH) that is suitable for massively parallel systems is proposed. The SBCH uses the hypercube topology as a basic building block and connects such building blocks by use of multidimensional spanning buses. In doing so, the SBCH 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 SBCH topology permits the efficient support of 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 SBCH network is its ability to support a large number of processors while maintaining a constant degree and a constant diameter. Other positive features include symmetry, incremental scalability, and fault tolerance. An optical implementation methodology is proposed for the SBCH. The implementation methodology combines the advantages of free-space optics with those of wavelength-division multiplexing techniques. An analysis of the feasibility of the proposed network is also presented.     

Experimental demonstration of the optical multi-mesh hypercube: scaleable interconnection network for multiprocessors and multicomputers

A prototype of a novel topology for scaleable optical interconnection networks called the optical multi-mesh hypercube (OMMH) is experimentally demonstrated to as high as a 150-Mbit/s data rate (2(7) - 1 nonreturn-to-zero pseudo-random data pattern) at a bit error rate of 10(-13)/link by the use of commercially available devices. OMMH is a scaleable network [Appl. Opt. 33, 7558 (1994); J. Lightwave Technol. 12, 704 (1994)] architecture that combines the positive features of the hypercube (small diameter, connectivity, symmetry, simple routing, and fault tolerance) and the mesh (constant node degree and size scaleability). The optical implementation method is divided into two levels: high-density local connections for the hypercube modules, and high-bit-rate, low-density, long connections for the mesh links connecting the hypercube modules. Free-space imaging systems utilizing vertical-cavity surface-emitting laser (VCSEL) arrays, lenslet arrays, space-invariant holographic techniques, and photodiode arrays are demonstrated for the local connections. Optobus fiber interconnects from Motorola are used for the long-distance connections. The OMMH was optimized to operate at the data rate of Motorola's Optobus (10-bit-wide, VCSEL-based bidirectional data interconnects at 150 Mbits/s). Difficulties encountered included the varying fan-out efficiencies of the different orders of the hologram, misalignment sensitivity of the free-space links, low power (1 mW) of the individual VCSEL's, and noise.     

Feasibility study of a scalable optical interconnection network for massively parallel processing systems

The theoretical modeling of a novel topology for scalable optical interconnection networks, called optical multimesh hypercube (OMMH), is developed to predict size, bit rate, bit-error rate, power budget, noise, efficiency, interconnect distance, pixel density, and misalignment sensitivity. The numerical predictions are validated with experimental data from commercially available products to assess the effects of various thermal, system, and geometric parameters on the behavior of the sample model. OMMH is a scalable network architecture that combines positive features of the hypercube (small diameter, regular, symmetric, and fault tolerant) and the mesh (constant node degree and size scalability). The OMMH is implemented by a free-space imaging system incorporated with a space-invariant hologram for the hypercube links and fiber optics to provide the mesh connectivity. The results of this work show that the free-space links can operate at 368 Mbits/s and the fiber-based links at 228 Mbits/s for a bit-error rate of 10(-17) per channel. The predicted system size for 32 nodes in the OMMH is 4.16 mm × 4.16 mm × 3.38 cm. Using 16-bit, bit-parallel transmission per node, the system can operate at a bit rate of up to 5.88 Gbits/s for a size of 1.04 cm × 1.04 cm × 3.38 cm.     

一种可扩展的双层光互连网络设计与分析

本文设计一种具有可扩展性的双层并行光互连网络.顶层为数字路由结点和光网络接口卡组成的星型网,底层为光网络接口卡连接而成的环形网.结点机以及数字路由结点影响网络的性能.结点机的吞吐能力限制了整个网络的吞吐率;扩展PCI总线的位数能够提高光网络接口卡的吞吐速率,采用64bit/66MHz工作模式可获得4.224 Gbps峰值传输速率.网络的实际最大吞吐速率为8.448Gbps,环网内平均延迟2195ns,环网间平均延迟4713 ns.可以采用本文设计的数字路由结点对网络进行低成本级联扩展,扩展后网络性能显著提高.     

Hierarchical optical ring interconnection (HORN): scalable interconnection network for multiprocessors and multicomputers

A new interconnection network for massively parallel computing is introduced. This network is called a hierarchal optical ring interconnection (HORN). The HORN consists of a single-hop, scalable, constant-degree, strictly nonblocking, fault-tolerant interconnection topology that uses wavelength-division multiple access to provide better utilization of the terahertz bandwidth offered by optics. The proposed optical network integrates the attractive features of hierarchical ring interconnections, e.g., a simple node interface, a constant node degree, better support for the locality of reference, and fault tolerance, with the advantages of optics. The HORN topology is presented, its architectural properties are analyzed, and an optical design methodology for it is described. Furthermore, a brief feasibility study of the HORN is conducted. The study shows that the topology is highly amenable to optical implementation with commercially available optical elements.     

Time multiplexing and control for optical cellular-hypercube arrays

We discuss the cellular-hypercube optical free-space interconnection architecture and its implementation by two-dimensional smart-pixel optoelectronic cellular arrays. We emphasize the behavior of the cellular hypercube in performing shift-invariant parallel shifts of data, a basic requirement of most single-instruction multiple-data algorithms. We present a time-multiplexing scheme for realizing the cellular hypercube, showing that the communication time is inversely proportional to the number of optical detectors per cell. We also present an improved hybrid interconnection network with improved performance that combines the cellular hypercube and mesh, using optics for the longer-distance connections and electronics for nearest-neighbor connections.     

Fault impact and fault tolerance in multiprocessor interconnection networks

Growing complexity of parallel machines coupled with increasing chip densities escalates the need for fault tolerance and recovery in these systems. In pursuit of the goal of fault-tolerant multiprocessors, many techniques have been proposed. Since methods for designing fault-tolerant processors and memories are relatively mature, the techniques considered in this paper focus on the interconnection network (ICN) linking the processors. The impact of faults on non-fault-tolerant ICNs is contrasted with that in fault-tolerant networks. Fault tolerance in ICNs is addressed at two levels, inter-node or switch level and system level. Inter-node or switch level pertains to data and control integrity and system level deals with maintaining network connectivity and adequate performance levels in the presence of faults. Fault-tolerant schemes at the switching element level warrant some form of concurrent error detection such as the use of codes usually combined with a full handshake protocol. Space–time trade-offs involved in the use of various codes and protocols are investigated. At the system level, several augmented multi-stage switching ICNs, tree and ring networks are studied. The combined provision for fault tolerance together with improved performance in the non-fault condition is emphasized. Finally, strategies for network reconfiguration and rerouting after system failure are presented.     

波长路由在并行计算机光互连中的应用

提出了一种波长路由并行光互连技术，其路由直接在源端用目标地址选择波长来建立，各波长有独立传输路径并采用光通道复用，路由变换节点为全光结构，光信号在变换节点处无转发延迟。此技术可应用于并行计算机互连网络和分布式高性能计算机群的互连网络。     

Modeling and analysis of fault tolerant multistage interconnection networks

Performance and reliability are two of the most crucial issues in today's high-performance instrumentation and measurement systems. High speed and compact density multistage interconnection networks (MINs) are widely-used subsystems in different applications. New performance models are proposed to evaluate a novel fault tolerant MIN arrangement, thereby assuring performance and reliability with high confidence level. A concurrent fault detection and recovery scheme for MINs is considered by rerouting over redundant interconnection links under stringent real-time constraints for digital instrumentation such as sensor networks. A switch architecture for concurrent testing and diagnosis is proposed. New performance models are developed and used to evaluate the compound effect of fault tolerant operation (inclusive of testing, diagnosis, and recovery) on the overall throughput and delay. Results are shown for single transient and permanent stuck-at faults on links and storage units in the switching elements. It is shown that performance degradation due to fault tolerance is graceful while performance degradation without fault recovery is unacceptable.     

Double-layer networks with holographic optical switches

The double-layer networks have the advantages of being strictly nonblocking and having a simpler routing algorithm, the lowest system insertion loss, a zero differential loss, fewer drivers, fewer interconnection lines, fewer crossovers, and the best signal-to-noise-ratio characteristic compared with any nondilated network. Using holographic optical switches to construct these networks not only eliminates all interconnection lines and crossovers but also reduces the number of drivers.     

Three-dimensional parallel unstructured grid generation

An algorithm for the parallel generation of 3-D unstructured grids is presented. The technique is an extension of the algorithm presented in Reference 21 for the 2-D case. The method uses a background grid as the means to separate spatially different regions, enabling the concurrent, parallel generation of elements in different domains and interdomain regions. The parallel 3-D grid generator was implemented and tested on the INTEL hypercube and Touchstone Delta parallel computers. The results obtained demonstrate the effectiveness of the algorithm developed. The methodology is applicable to the parallel implementation of a wide range of problems that are, in principle, scalar by nature, and do not lend themselves to SIMD parallelization.     

Compatibility of TCP Reno and TCP Vegas in wireless ad hoc networks

Previous work has shown that TCP (transmission control protocol) Vegas outperforms the more widely deployed TCP Reno in both wired and wireless networks. It was also shown that when both TCP variants coexist on the same wired links, Reno dominates because of its more aggressive behaviour. This paper examines for the first time the compatibility between Reno and Vegas in wireless IEEE 802.11 ad hoc networks. It is shown that Vegas generally dominates in the heterogeneous Reno/Vegas network scenario; a startling result that is inconsistent with what is seen in wired networks. It is shown that the wireless ad hoc network environment does not reward the aggressive behaviour of Reno. On the other hand, Vegas, with its more accurate yet more conservative mechanisms, is able to capture most of the bandwidth. This is found to be true when using the on-demand routing protocols of dynamic source routing (DSR) or ad hoc on-demand distance vector (AODV): the failure of a node to reach a next-hop node because of media access control (MAC)-sublayer repeated collisions is reported to the routing protocol, which then declares a route error that impacts Reno in a more serious way than Vegas. When the table-driven routing protocol destination-sequenced distance vector (DSDV) is used, Reno and Vegas share the network bandwidth in a fairer manner. Generally, fairness in this environment can be improved by reducing the TCP maximum window size.     

Performance evaluation of shortest multipath source routing scheme

Multipath routing in mobile ad-hoc networks allows the establishment of multiple paths for routing between a source-destination pair. It exploits the resource redundancy and diversity in the underlying network to provide benefits such as fault tolerance, load balancing, bandwidth aggregation and the improvement in quality-of-service metrics such as delay. Previous work shows that on-demand multipath routing schemes achieve better performance under certain scenarios with respect to a number of key performance metrics when compared with traditional single-path routing mechanisms. A multipath routing scheme, referred to as shortest multipath source (SMS) routing based on dynamic source routing (DSR) is proposed here. The mechanism has two novel aspects compared with other on-demand multipath routing schemes: it achieves shorter multiple partial-disjoint paths and allows more rapid recovery from route breaks. The performance differentials are investigated using NS-2 under conditions of varying mobility, offered load and network size. Results reveal that SMS provides a better solution than existing source-based approaches in a truly mobile ad-hoc environment.     

Reactive and proactive routing in labelled optical burst switching networks

Optical burst switching architectures without buffering capabilities are sensitive to burst congestion. The existence of a few highly congested links may seriously aggravate the network throughput. Proper network routing may help in congestion reduction. The authors focus on adaptive routing strategies to be applied in labelled OBS networks, that is, with explicit routing paths. In particular, two isolated alternative routing algorithms that aim at network performance improvement because of reactive route selection are studied. Moreover, a nonlinear optimisation method for multi-path source-based routing, which aims at proactive congestion reduction is proposed. Comparative performance results are provided and some implementation issues are discussed.     

一种结合网络编码的路径代价衡量方法

针对无线路由协议中的路径代价衡量问题,结合网络编码改善无线节点信息互换的思想,提出了一种结合网络编码的路径代价衡量方法--RMNC,其核心思想是利用流量参数反映信息流的网络编码"搭乘"程度和逐节点计算路径的代价.通过将传输流流量参数和路径中节点左右链路信息流流量参数进行运算,获得路径上的各个节点的传输代价;网络中某一条路径的代价等于组成这条路径的节点传输代价之和,通过比较不同路径的逐节点计算代价值,获得最短路径.分析和模拟测试结果表明,RMNC可以有效地获得结合网络编码的最短路径,达到提高传输性能的目的.尽管传输延时有所增加,但可以接受,方法可行.     

Generalized methodology for modeling and simulating optical interconnection networks using diffraction analysis

Research in the field of free-space optical interconnection networks has reached a point where simulators and other design tools are desirable for reducing development costs and for improving design time. Previously proposed methodologies have only been applicable to simple systems. Our goal was to develop a simulation methodology capable of evaluating the performance characteristics for a variety of different free-space networks under a range of different configurations and operating states. The proposed methodology operates by first establishing the optical signal powers at various locations in the network. These powers are developed through the simulation by diffraction analysis of the light propagation through the network. After this evaluation, characteristics such as bit-error rate, signal-to-noise ratio, and system bandwidth are calculated. Further, the simultaneous evaluation of this process for a set of component misalignments provides a measure of the alignment tolerance of a design. We discuss this simulation process in detail as well as provide models for different optical interconnection network components.     

Bit-level packet-switching all-optical multihop shuffle networks with deflection routing

We propose an all-optical packet-switching scheme in multihop shuffle networks in which deflection routing is used as its contention-resolution principle. In our scheme only partial address information in the packet header is read before a routing decision is made. Because the new scheme does not involve a time-consuming look-up table, extremely low latency operation is possible at each node. Moreover, because the number of demultiplexers at each node can be kept constant even though the network size changes, cost-effective design of a node is possible.     

On Generalized Fibonacci Cubes and Unitary Transforms

 We present a new interconnection topology called generalized Fibonacci topology, which unifies a wide range of connection topologies such as the Boolean cube (or hypercube), classical Fibonacci cube, etc. Some basic topological properties of generalized Fibonacci cubes are established. Finally, we developed new classes of the discrete orthogonal transforms, based on the generalized Fibonacci recursions. They can be implemented efficiently by butterfly-type networks (like the Fourier, or the Haar transforms). A generalized Fibonacci cube based processor architecture (generalizing the known SIMD architecture — hypercube processor) can be efficiently used for hardware implementation of the proposed discrete orthogonal transforms. Received: October 31, 1996     

Optical implementation of rearrangeable nonblocking double banyan interconnection network in free space

The banyan network plays an important role in optical interconnection networks. A smart and compact double banyan network with cascading banyan network and inverse banyan network is proposed by using a polarizing beam-splitter (PBS), a phase spatial light modulator (PSLM), a half-wave plate (HWP), a double-faced reflective mirror (DFRM), and mirrors. The PBS features that the s-component (perpendicular to the incident plane) of the incident light beam is reflected, and the p-component (parallel to the incident plane) passes through it. Simultaneously, the bipartition graph algorithm (BGA) is adopted to ascertain the state of the node switch in each node stage (straight or crossover connection). According to switching logic, under control of external electrical signals, the PSLM functions to control routing paths of the signal beams, i.e. the polarization of each optical signal is rotated or not rotated 90° by a programmable PSLM. Since the proposed optical setup consists of only optical polarization elements, it is compact in structure, and possesses a low energy loss, a high signal-to-noise ratio and an available large number of optical channels. Finally, the discussions and the experimental results show that the double banyan network proposed here, owing to without signal blocking and conflict, may be used in optical communication and optical information processing.