Optical Interconnection Networks Research Articles

Integrated fault tolerant connections-scheduling for dilated Benes network

The 2×2 electro-optical directional coupler is a common switching element in dilated optical multistage interconnection networks (DOMINs) and can be applied to photonic switching technology. Because of the undesirable coupling within an optical switch, a portion of the energy of the input signal might be transmitted to the undesired output terminal in the switch. A switch is said to be faulty if it causes excessive crosstalk above a given threshold, and optical crosstalk is a problem commonly found in DOMINs. The problem of connections-scheduling with faulty switches in a dilated re-arrangeable photonic network, say dilated Benes network (DBN), is studied and an integrated fault tolerant DBN architecture is proposed. Based on the faulty switch information given from the fault diagnosis algorithm, the fault-free path generation algorithm and scheduling algorithm are devised to schedule a set of connections avoiding faulty switches with a polynomial time complexity in the DBN. It is desirable to keep the number of rounds as small as possible. The simulation results show that the proposed method outperforms the greedy-scheduling algorithm in terms of routing number, number of failed connections, minimum number of rounds, and connection success probability when the size of the network is large. Furthermore, the proposed methodology can be embedded in the optical switching system to achieve self-testing.

Optimal all-to-all personalised exchange in a novel optical multistage interconnection network

An all-to-all personalised exchange is one of the most dense collective communication operations in parallel and distributed computing and communication applications. Each node in the network needs to send a different message to each of the other nodes. Advances in electro-optic switches have made optical communication a good networking choice that can satisfy the high channel bandwidth, low communication latency, low error rate, and gigabit transmission requirements of high performance computing and communication applications. Previously proposed optical multistage interconnection networks (MINs) require at least two passes to send a message from each node to a different node (to realise a permutation) in the network. In this paper, we propose an MIN that requires only one pass to realise a permutation. That is, the new MIN requires n 1 passes instead of 2n passes, the requirement for other optical MINs, for an all-to-all personalised exchange. The new network is optimal in terms of the number of passes that is required for an all-to-all personalised exchange.

An optical interconnection network and a modified snooping protocol for the design of large-scale symmetric multiprocessors (SMPs)

In symmetric multiprocessors (SMPs), the cache coherence overhead and the speed of the shared buses limit the address/snoop bandwidth needed to broadcast transactions to all processors. As a solution, a scalable address subnetwork called symmetric multiprocessor network (SYMNET) is proposed in which address requests and snoop responses of SMPs are implemented optically. SYMNET not only uses passive optical interconnects that increases the speed of the proposed network, but also pipelines address requests at a much faster rate than electronics. This increases the address bandwidth for snooping, but the preservation of cache coherence can no longer be maintained with the usual snooping protocols. A modified coherence protocol, coherence in SYMNET (COSYM), is introduced to solve the coherence problem. COSYM was evaluated with a subset of Splash-2 benchmarks and compared with the electrical bus-based MOESI protocol. The simulation studies have shown a 5-66 percent improvement in execution time for COSYM as compared to MOESI for various applications. Simulations have also shown that the average latency for a transaction to complete using COSYM protocol was 5-78 percent better than the MOESI protocol. It is also seen that SYMNET can scale up to hundreds of processors while still using fast snooping-based cache coherence protocols, and additional performance gains may be attained with further improvement in optical device technology.

Optical interconnection networks: introduction to the feature issue

Feature Issue on Optical Interconnection Networks (OIN). We have assembled a feature issue of the Journal of Optical Networking on the exciting emerging field of Optical Interconnection Networks that includes contributions addressing a broad range of current research.

Emulation of realistic network traffic patterns on an eight-node data vortex interconnection network subsystem

Feature Issue on Optical Interconnection Networks (OIN). We demonstrate the feasibility of the data vortex interconnection network architecture for use in supercomputing by emulating realistic network traffic on an eight-node subnetwork. The evaluation workload uses memory accesses from the Barnes-Hut application in the SLPASH-2 parallel computing benchmark suite, which was extracted by using the M5 multiprocessor system simulator. We confirm that traffic is routed correctly and efficiently.

Performance evaluation of an optically interconnected "scheduling" switch network for Pareto traffic

Feature Issue on Optical Interconnection Networks (OIN). A performance analysis of an optically interconnected packet-scheduling switch network is presented. The scheduling switch uses a branch of feed-forward delays for each input port, interconnected with elementary optical switches to resolve contention. The scheduling switch is guaranteed to be lossless under a certain smoothness property condition. We investigate the packet-loss performance of the switch when the smoothness property condition does not hold, as well as the packet delay impairments at the edge of a scheduling switch interconnected network when this property is enforced.

All-optical signal processing for optical packet switching [Invited

Feature Issue on Optical Interconnection Networks (OIN). We present three optical signal processing functional blocks that enable 1×N optical packet switching. An ultrafast asynchronous multioutput all-optical header processor is demonstrated with a terahertz optical asymmetric demultiplexer in combination with a header preprocessor. It is shown that self-induced polarization rotation can be used for both the header processor and the header preprocessor. The second functional block is optical buffering. This is shown with both a laser neural network and a recirculating buffer. Related to this is a three-state all-optical memory based on coupled lasers, which increases the number of possible output states of an optical packet switch.

Optical-packet-switched interconnect for supercomputer applications [Invited

Feature Issue on Optical Interconnection Networks (OIN). We describe a low-latency, high-throughput scalable optical interconnect switch for high-performance computer systems that features a broadcast-and-select architecture based on wavelength- and space-division multiplexing. Its electronic control architecture is optimized for low latency and high use. Our demonstration system will support 64 nodes with a line rate of 40 Gbit/s per node and operate on fixed-length packets with a duration of 51.2 ns using burst-mode receivers. We address the key system-level requirements and challenges for such applications.

Ultrafast optoelectronic packet processing for asynchronous, optical-packet-switched networks [Invited

Feature Issue on Optical Interconnection Networks (OIN). We describe hybrid optical-electrical systems that perform header processing and buffering of ultrafast, asynchronous optical packets. Our systems are enabled by three key, novel devices: an all-optical serial-to-parallel converter, an optical clock-pulse generator, and a photonic parallel-to-serial-converter. These devices allow utilization of complementary metal-oxide semiconductor technology for compact, highly functional optical packet processing. A simplified node architecture for asynchronous, optical- packet-switched networks is made possible by these systems with all the necessary node functions integrated compactly. We also demonstrate an optical label swapper and a photonic random access memory for 40-Gbit/s, 16-bit, asynchronous optical packets.

Cyclic crossbar network for photonic switching

Feature Issue on Optical Interconnection Networks (OIN). The rectangular crossbar network is commonly used in switching networks because it has the advantages of easy connection-path rebuilding, wide-sense nonblocking, and no crossover in its interconnection lines. However, this kind of network has two major disadvantages: one is that it has nonzero differential loss, and the other is that its signal-to-noise ratio (SNR) is decreased by both the increase of the network dimension and the insertion loss of switches. Here the cyclic crossbar network structure is presented. The cyclic crossbar network is similar to the rectangular crossbar network. For both types of network, an N×N network is constructed by N columns, and each column consists of N switches. The major difference is that construction with N backslash lines and N slash lines is used for the cyclic crossbar network. For every backslash line to have its crosspoint on the right side, the network structure is designed into a circle shape. In this type of network, the differential loss is eliminated and SNR is independent on insertion loss of switches, but the original advantages are maintained.

Packet loss rate and jitter differentiating quality-of-service schemes for asynchronous optical packet switches

Feature Issue on Optical Interconnection Networks (OIN). We propose access-restriction-based quality-of-service differentiation schemes, suitable for an asynchronous optical packet switch with a contention resolution pool that contains both tunable wavelength converters and fiber delay lines. The schemes aim at obtaining a high degree of packet-loss-rate isolation, for a low increase in overall packet loss rate, at the same time respecting the jitter tolerance of each class of service. Numerical simulations quantify how the performance depends significantly on the jitter tolerance of the traffic in general, and of the highest-priority class of service in particular.

Priority scheme for supporting quality of service in optical burst switching networks

Feature Issue on Optical Interconnection Networks (OIN). We present what we believe to be a new scheme, called preemptive prioritized just enough time (PPJET), for quality-of-service (QoS) provisioning in bufferless optical burst switching (OBS) networks. PPJET provides better service for high-priority traffic by dropping reservations belonging to lower-priority traffic with a new channel-scheduling algorithm called preemptive latest-available unused channel with void filling (PLAUC-VF). An approximate method for calculating the dropping probability in PPJET is also discussed. Moreover, we perform discrete-event simulations to evaluate the performance of PPJET. Simulation results show that PPJET outperforms offset-based QoS schemes both in terms of dropping probability and end-to-end delay.

Investigation of the scalability of dynamic wavelength-routed optical networks [Invited

Feature Issue on Optical Interconnection Networks (OIN). We describe results of the scalability analysis for dynamic wavelength-routed optical networks with end-to-end lightpath assignment and central network control with electronic scheduling and processing of lightpath requests. We investigate the effect of the algorithm complexity in both the scheduling and the dynamic routing and wavelength assignment (DRWA) of lightpath requests. Scheduling theory and static performance-prediction techniques were applied to define the bounds on the electronic processing time of requests, and hence the maximum number of nodes supported by a centralized dynamic optical network for given blocking probability, latency, and network diameter. Scalability analysis results show that medium-sized centralized networks (~50 nodes) can be supported when these networks are reconfigured on a burst-by-burst basis. In addition, we found that real topologies showed a complex trade-off between the request processing time, blocking probability, and resource requirements. These findings can be used to determine the optimum combination of scheduling/DRWA algorithm, showing that the fastest DRWA algorithm does not necessarily lead to the minimum blocking probability and maximum scalability but that a careful consideration of both blocking and processing speed is required. The results are applicable both to dynamic network architectures with centralized request processing such as wavelength-routed optical networks and to the design of advanced optical switching matrices and routers.

Design alternatives for optical-packet-interconnection network architectures [Invited

Feature Issue on Optical Interconnection Networks (OIN). In the design of an optical-packet-switching network, several policy decisions must be made and network parameters must be dimensioned. These decisions determine the network operation and performance. We present an overview of the design alternatives for optical-packet-interconnection network architectures. The issues that are discussed include contention resolution schemes, header processing, and quality-of-service support. The design of metropolitan-area packet-switched networks is also presented in a separate section.

Alternative structures for two-dimensional MEMS optical switches [Invited

Feature Issue on Optical Interconnection Networks (OIN). Two-dimensional (2-D) microelectromechanical system (MEMS) optical switches have the merits of easy fabrication and high reliability. Since the optical signal loss is mainly proportional to the length of signaling paths in the switches, current 2-D MEMS optical switches that use a crossbar structure have a rather limited number of ports. For larger 2-D MEMS optical switches, we may use nonrectangular topology switching fabrics to shorten the internal signaling path or to recollimate the optical signal segment by segment inside the switches. We discuss these approaches from the aspect of implementation and routing control complexity.

Implementation of medium access control for interconnecting slotted rings to form a WDM metropolitan area network

Feature Issue on Optical Interconnection Networks (OIN). A system of slotted interconnected rings employing a combination of wavelength-division multiple access (WDMA) with time-division multiple access (TDMA) can serve a metropolitan area without electro-optical conversion and buffering of payload except at system entry points. The multiple rings overcome the power budget limitations of the single ring extending the reach of the system to even the largest metropolitan areas, the WDM dimension provides flexibility and ease of evolution, and the TDMA dimension offers the efficiency of multiplexing gain particularly under bursty traffic. The system control information is transferred on a dedicated wavelength and is processed in the electrical domain at the ring nodes and the hub, which interconnects the rings. The algorithms control the access to each ring and the scheduling of slots among the rings, based on explicit reservations, to adapt efficiently to the fluctuating offered load. We present the design and hardware implementation of the access control algorithms for such a system built in the framework of the Information Society Technologies (IST) project DAVID (data and voice over DWDM).

Diagnosing crosstalk faults algorithm design and analysis in a class of dilated re-arrangeable photonic switching networks

Optical multistage interconnection networks (OMINs), based on 2 × 2 directional coupler photonic switches, play a significant role in all-optical lightwave networks. Space dilation reduces the native effect of switch crossover. This paper presents two types of dilated re-arrangeable networks extended from dilated blocking networks: a class of two cascaded dilated blocking networks, such as dilated_ blocking_ network-dilated_ blocking_ network architectures, and the other type can be extended from the dilated blocking networks, i.e. horizontal cascading (HC) or vertical stacking (VS). Optical crosstalk, an underestimated problem in OMINs, is the effect of undesirable coupling of signals between paths within a directional coupler due to its rapid aging. A switch is faulty if it causes excessive crosstalk in excess of a given threshold. Diagnosing faulty switch is essential issue in OMINs. Previous work [Inform. Sci.: Int. J. 124 (2000) 59] has developed the appropriate diagnostic algorithm for dilated blocking photonic networks. This paper devises diagnostic algorithms for dilated re-arrangeable networks, which benefits from [Inform. Sci.: Int. J. 124 (2000) 59]. The proposed algorithm is also verified in this paper.

Θ(N(superscript 5/6)) Time Complexity of Fault Diagnosis Algorithms in N×N Dilated Blocking Photonic Switching Networks

Dilated Optical Multistage Interconnection Networks (DOMINs) based on 2×2 directional coupler photonic switches play an important role in all-optical high-performance networks, especially for the emerging IP over DWDM architectures. The problem of crosstalk within photonic switches is underestimated due to the aging of the switching element, control voltage, temperature and polarization, and thus causes undesirable coupling of the signal from one path to the other. Previous works [18] designed an efficient diagnosing disjoint faults algorithm in small sized networks, which reduced the number of tests required by overlapping the tests with computations to one half in photonic switching networks. Furthermore, this paper generically derives algorithms and mathematical modules to find the optimal degree of parallelism of faults diagnosis for N×N dilated blocking networks, as the size of network is larger. Taking advantage of the properties of disjoint faults, diagnosis can be accelerated significantly because the optimal degree of parallel fault diagnosis may grow exponentially. To reduce the diagnosis time, an algorithm is proposed herein to find the maximum number of disjoint faults. Rather than requiring up to 4MN tests as a native approach, a two-phase diagnosis algorithm is proposed to reduce the testing requirement to 4N tests. This study extends the concept of disjoint faults to reduce the number of tests to the time efficiency of Θ(N(superscript 5/6)) for N×N DOMINs.

SYMNET: an optical interconnection network for scalable high-performance symmetric multiprocessors.

We address the primary limitation of the bandwidth to satisfy the demands for address transactions in future cache-coherent symmetric multiprocessors (SMPs). It is widely known that the bus speed and the coherence overhead limit the snoop/address bandwidth needed to broadcast address transactions to all processors. As a solution, we propose a scalable address subnetwork called symmetric multiprocessor network (SYMNET) in which address requests and snoop responses of SMPs are implemented optically. SYMNET not only has the ability to pipeline address requests, but also multiple address requests from different processors can propagate through the address subnetwork simultaneously. This is in contrast with all electrical bus-based SMPs, where only a single request is broadcast on the physical address bus at any given point in time. The simultaneous propagation of multiple address requests in SYMNET increases the available address bandwidth and lowers the latency of the network, but the preservation of cache coherence can no longer be maintained with the usual fast snooping protocols. A modified snooping cache-coherence protocol, coherence in SYMNET (COSYM) is introduced to solve the coherence problem. We evaluated SYMNET with a subset of Splash-2 benchmarks and compared it with the electrical bus-based MOESI (modified, owned, exclusive, shared, invalid) protocol. Our simulation studies have shown a 5-66% improvement in execution time for COSYM as compared with MOESI for various applications. Simulations have also shown that the average latency for a transaction to complete by use of COSYM protocol was 5-78% better than the MOESI protocol. SYMNET can scale up to hundreds of processors while still using fast snooping-based cache-coherence protocols, and additional performance gains may be attained with further improvement in optical device technology.

Permutation routing in optical MIN with minimum number of stages

In a hybrid optical multistage interconnection network (MIN), optical signals are routed by electronically controlled switches using directional couplers. A relevant design problem is to minimize the path-dependent loss of the optical signal, which is directly proportional to the number of couplers, i.e., the number of switches through which the signal has to pass. In general, given the network size and the type of the MIN, the number of stages is a constant. Hence, an input signal has to pass through a fixed number of couplers to reach the output.In this paper, it is shown that the routing delay and path-dependent loss in a fixed-stage N × N MIN, can be significantly reduced on the average by using a variable-stage shuffle-exchange network instead. An arbitrary N × N permutation P can be routed with minimum delay and minimum path-dependent loss, if the minimum number of stages of the MIN necessary to route P is known. An O(Nn) algorithm (N = 2n) is presented here for checking the admissibility of a given permutation P in an m-stage shuffle-exchange network (SEN), where 1 ≤ m ≤ n. The minimum-stage SEN needed to pass P can then be determined in O(Nn log n) time. Furthermore, for n ≤ m ≤ 2n - 1, a necessary condition for permutation admissibility is derived which is shown to be necessary as well as sufficient for the special class of BPC (bit-permute-complement) permutations. It has been shown that, for 1 ≤ m ≤ 2n - 1, the minimum number of stages required to pass a BPC permutation P through a SEN can be determined in O(n2) time, and P can be routed through a variable-stage SEN using the minimum number of stages only. In an optical MIN, this technique helps to reduce the path-dependent loss by limiting the number of stages to be traversed by the optical signal.