Optical Interconnection Networks Research Articles

The g-Good-Neighbor Conditional Diagnosabilities of Hypermesh Optical Interconnection Networks Under the PMC and Comparison Models

Processor fault diagnosis is intended to identify faulty processors of a multicomputer system, guaranteeing its high reliability and availability. The [Formula: see text]-good-neighbor conditional diagnosability is a novel fault diagnosis method for various networks. Under the PMC model and comparison model, the [Formula: see text]-good-neighbor conditional diagnosabilities of hypermesh optical interconnection networks are determined, respectively. Directly applying the results, the [Formula: see text]-good-neighbor conditional diagnosabilities of hypercubes are derived under the PMC and comparison models.

Flat Ball: Dynamic topology for energy management of optical interconnection networks in data centers

Cloud computing and web applications have further highlighted the need for powerful data centers with large bandwidth and low power consumption. In order to provide such a large bandwidth, today's electronic data centers require high power levels. Furthermore, communications and devices are not always used to their full potential. Optical networks can be a proper option for data centers, as they can offer large and variable bandwidth with lower power consumption compared to their electronic counterparts. In this regard, a dynamic all-optical network architecture is proposed in this paper, with bandwidth reconfiguration capability and low latency to reduce energy consumption. This proposed architecture, called Flat Ball, offers dynamic bandwidth utilizing passive optical devices with low power consumption and latency. Under realistic data center traffic scenarios, the latency of this network is up to 50% lower than its electrical counterpart, while providing a much higher throughput and consuming up to 60% less power.

A machine learning assisted optical multistage interconnection network: Performance analysis and hardware demonstration

AbstractIntegration of the machine learning (ML) technique in all‐optical networks can enhance the effectiveness of resource utilization, quality of service assurances, and scalability in optical networks. All‐optical multistage interconnection networks (MINs) are implicitly designed to withstand the increasing high‐volume traffic demands at data centers. However, the contention resolution mechanism in MINs becomes a bottleneck in handling such data traffic. In this paper, a select list of ML algorithms replaces the traditional electronic signal processing methods used to resolve contention in MIN. The suitability of these algorithms in improving the performance of the entire network is assessed in terms of injection rate, average latency, and latency distribution. Our findings showed that the ML module is recommended for improving the performance of the network. The improved performance and traffic grooming capabilities of the module are also validated by using a hardware testbed.

128 Gbit/s high speed optical interconnection networks in data centers by a 30 GHz Mach-Zehnder modulator

128 Gbit/s high speed optical interconnection networks in data centers by a 30 GHz Mach-Zehnder modulator

Efficient mode converters and filters using asymmetrical directional couplers with subwavelength gratings.

Mode-division multiplexing (MDM) is a promising solution to improve data transmission capacity for future optical interconnect networks. Mode converters and mode filters play a key role in on-chip MDM systems. Here, we propose and experimentally demonstrate a device, enabling mode conversion and filtering simultaneously, which is composed of asymmetrical directional couplers with subwavelength gratings, in a small footprint of 14.7 µm × 1.42 µm. The device can realize optical mode conversion between the first-order transverse electric (TE) mode and the fundamental TE mode, and can also filter the fundamental TE mode efficiently. The conversion efficiency is over 95%, with a broad 1 dB bandwidth over 80 nm and a high mode extinction ratio of >29 dB. As a mode filter, strong mode elimination of >30 dB is achieved.

SDM-WDM ROADM Based on Direct-Laser-Writing SDM Core Separator

Weakly-coupled space division multiplexing (SDM) techniques based on multi-core fibers (MCF) have attracted great attention for the capacity enhancement of optical interconnect systems and networks. In this paper, we propose a novel direct-laser-writing (DLW) core separator device and apply it in a reconfigurable optical add/drop multiplexer (ROADM) structure for two-dimensional SDM and wavelength division multiplexing (WDM) switching. The proposed core separator matching an 8-core MCF is fabricated. Then assisted with MCF rotators, independent adding and dropping functionalities for arbitrary core are experimentally demonstrated with slight power penalty for both core and wavelength dimensions. The influence of crosstalk for cascading core separators is also investigated and below 0.35-dB receiving sensitivity penalty is obtained when three cores are dropped one by one. The compact DLW core separators can reduce the complexity of SDM nodes and will be beneficial for high-speed optical interconnects.

Metal-printing tunable interlayer waveguide coupler using low-loss fluorinated polycarbonate.

Tunable three-dimensional (3D) integrated optical waveguide chips with optical interconnection function are beneficial to expand the application of optical devices in a 3D integrated photonic module. Here, we propose a thermo-optic (TO) tunable interlayer waveguide coupler based on the metal-printing technique. Low-loss fluorinated polycarbonate (AF-Ali-PC MA) and poly (methyl methacrylate-glycidyl methacrylate) [P(MMA-co-GMA)] are synthesized as waveguide core and cladding layer, respectively. The thermal stability and optical adsorption characteristics of AF-Ali-PC MA are analyzed. Optical signal transmission features of the interlayer coupling waveguides are simulated. The optical response properties and fabrication process flows of a dynamic multilayer waveguide chip can be greatly improved by the metal-printing technique. The on-off time of the TO interlayer coupling chip is obtained as 250 µs, and the electrical power consumption is measured as 7.6 mW. To the best of our knowledge, this is the first time that a TO tunable interlayer waveguide coupler is achieved by an efficient metal-printing method, which is suitable for large-scale photonic integrated circuit (PIC) systems and multilayer optical interconnection (OXC) networks.

Study on Topology Reconfiguration of Optical Interconnection Network in Data Center

Study on Topology Reconfiguration of Optical Interconnection Network in Data Center

Anycast Service Grooming Algorithm of Cloud Computing Based on Wireless Communication Network

In order to improve the energy efficiency of optical interconnection network between data centers, a hybrid path transmission energy-saving routing algorithm with service duration awareness is proposed for Anycast service. In order to reduce the number of new optical path and working components, single path transmission is preferred; If the traffic is blocked, the transmission energy consumption is minimized. In addition, the traffic grooming strategy based on spectrum reservation is introduced to reduce the cost of protecting bandwidth and optical transceiver. Simulation results show that: compared with the traditional energy-saving routing algorithm, the proposed algorithm can significantly reduce the network energy consumption, effectively avoid the excessive increase of traffic blocking rate, and achieve the balance of network energy consumption and performance.

Fault Tolerance of Optical Hypercube Interconnection Networks with r -Communication Pattern

As power grids and optical interconnection networks are interdependent, the reliabilities of the optical networks are critical issues in power systems. The optical networks hold prominent performance including wide bandwidth, low loss, strong anti-interference capability, high fidelity, and reliable performance. They are regarded as promising alternatives to electrical networks for parallel processing. This paper is aimed at taking the first step in understanding the communication efficiencies of optical networks. For that purpose, on optical networks, we propose a series of novel notions including communication pattern, r -communication graph, reduced diameter, enhanced connectivity, r -diameter, and r -connectivity. Using these notions, we determine that the r -diameter and r -connectivity of the optical n -dimensional hypercube network are n / r and n 1 + n 2 + ⋯ + n r , respectively. Since the parameter r is variable, we can adjust different values of r on the basis of the wavelength resources and load of the optical networks, achieving enhanced communication efficiencies of these networks. Compared with the electric n -dimensional hypercube network, the proposed communication pattern on the optical hypercube network not only reduces the maximum communication delay of the conventional electrical hypercube significantly but also improves its fault tolerance remarkably.

Inverse-designed ultra-compact high efficiency and low crosstalk optical interconnect based on waveguide crossing and wavelength demultiplexer

In this paper, we use the inverse design method to design an optical interconnection system composed of wavelength demultiplexer and the same direction waveguide crossing on silicon-on-insulator (SOI) platform. A 2.4 μm × 3.6 μm wavelength demultiplexer with an input wavelength of 1.3–1.6 μm is designed. When the target wavelength of the device is 1.4 μm, the insertion loss of the output port is − 0.93 dB, and there is − 18.4 dB crosstalk, in TE0 mode. The insertion loss of the target wavelength of 1.6 μm in TE0 mode is − 0.88 dB, and the crosstalk is − 19.1 dB. Then, we designed a same direction waveguide crossing, the footprint is only 2.4 μm × 3.6 μm, the insertion loss of the wavelength 1.4 μm and 1.6 μm in TE0 mode is − 0.99 dB and − 1 dB, and the crosstalk is − 12.14 dB and − 14.34 dB, respectively. Finally, an optical interconnect structure composed of two devices is used, which can become the most basic component of the optical interconnect network. In TE0 mode, the insertion loss of the output wavelength of 1.4 μm at the output port is − 1.3 dB, and the crosstalk is − 29.36 dB. The insertion loss of the output wavelength of 1.6 μm is − 1.39 dB, and the crosstalk is − 38.99 dB.

FPGA-based implementation of two-step schedulers for modular optical interconnection networks

Optical interconnection networks promise to overcome the limitations of current electronic switching fabrics, enabling higher throughput, lower latency, and lower power consumption. Multi-plane architectures, based on multiple optical switching domains (e.g., space, time, wavelength, orbital angular momentum), are gaining research attention because of their modularity and scalability compared to single-domain switches. An effective scheduler, namely, the two-step scheduler (TSS), has been proposed for multi-plane optical interconnection networks, exploiting their modularity to speed up computations while satisfying the peculiar scheduling constraints. In this paper, a hardware implementation of TSS for modular optical interconnection networks is presented and thoroughly assessed. Both scheduling steps are parallelized with the aim of optimizing the execution time. iSLIP and longest queue first (LQF) scheduling algorithms are exploited in each step, resulting in four TSS configurations that are compared among each other and with classical single-step schedulers (SSSs) in terms of scheduling and hardware performance. TSS outperforms SSS in terms of the number of iterations, maximum operating frequency, worst-case scheduling duration, and required logic resources (i.e., scalability) at the expense of a slight latency penalty. Among all TSS configurations, LQF-based TSS guarantees the lowest scheduling latency, while iSLIP-based TSS minimizes the scheduling duration and the use of field programmable gate array (FPGA) resources.

Analysis of phase estimation for star QAM coherent optical packet switched system

The development of data centers, 5th generation wireless network remote network education and cloud computing is increasingly demanding optical packet switched (OPS) optical interconnect networks. The blind phase estimation algorithm is one of the important phase estimation algorithms in the star QAM OPS network. However, the blind phase estimation algorithm can cause system performance degradation rapidly in the case of low signal to noise ratio (SNR). To our best knowledges, the principle and processing mechanism of the problem have not been analyzed and studied in detail. In this paper, the blind phase estimation is analyzed for the first time based on SNR and phase noise probability model at low optical signal to noise ratio. Based on the analysis, the common single differential and doubly differential coding schemes are explained for blind phase estimation. Then, the performance of the blind phase estimation + single differential and blind phase estimation + doubly differential algorithms for star QAM OPS system be compared in detail, including time-varying phase processing capability, transmission performance and tolerance to laser linewidth. The experiments be performed to verify the performance of the blind phase estimation + single differential and blind phase estimation + doubly differential algorithms.

FOSquare: A Novel Optical HPC Interconnect Network Architecture Based on Fast Optical Switches with Distributed Optical Flow Control

Interconnecting networks adopting Fast Optical Switches (FOS) can achieve high bandwidth, low latency, and low power consumption. We propose and demonstrate a novel interconnecting topology based on FOS (FOSquare) with distributed fast flow control which is suitable for HPC infrastructures. We also present an Optimized Mapping (OPM) algorithm that maps the most communication-related processes inside a rack. We numerically investigate and compare the network performance of FOSquare with Leaf-Spine under real traffic traces collected by running multiple applications (CG, MG, MILC, and MINI_MD) in an HPC infrastructure. The numerical results show that the FOSquare can reduce >10% latency with respect to Leaf-Spine under the scenario of 16 available cores.

New activation functions for single layer feedforward neural network

Artificial Neural Network (ANN) is a subfield of machine learning and it has been widely used by the researchers. The attractiveness of ANNs comes from their remarkable information processing capability. An important key parameter of ANN is activation function (AF). Different AFs can significantly affect the performance of an ANN, and therefore choosing a good AF is important. This study aims to introduce some new AFs that combine the advantages of predefined AFs and perform better than them. For this purpose, we propose some new AFs, which we called generalized swish, mean-swish, ReLU-swish, triple-state swish, sigmoid-algebraic, triple-state sigmoid, exponential swish, sinc-sigmoid and derivative of sigmoid AFs. Then, we compare the proposed AFs with some well-known and recently proposed AFs. To investigate the performance of these AFs, we use four different data sets, which are simulated data, optical interconnection network data, specifications of cars, and average house costs data.

Modeling and Performance Enhancement of Active Hybrid Plasmonic Electro-optic Routing Switch

This paper produces a more compact and efficient design for hybrid plasmonic 2 × 2 electro-optic routing switch, with aid of an active material that has an electrically tunable permittivity, and discusses its ability for realizing more advanced structures. The proposed 2 × 2 electro-optic switch, with foot print of 11.6 μm2 and switching time of 0.949 pS, satisfies transmission values (coupling efficiency) less than 7% at OFF state and greater than 75% at ON state at the telecommunication wavelength. This design is used as a base unit to realize two advanced structures: (a) the 3 × 3 switching matrix, with a foot print of 47–49 μm2, which satisfies transmission greater than 60% at ON states and transmission less than 10% at OFF states, and (b) the 4 × 4 switching matrix, with a foot print of 88 μm2, which satisfies transmission greater than 70% at ON states and transmission less than 5% at OFF states. The obtained results have a validation with older researches. A 3D finite-difference time-domain (FDTD) solver is employed to simulate and characterize the realized structures. The components would be useful in the optical interconnect networks, photonic integrated circuits, and signal processing system.

Research on Topoloty Reconstruction Mechanism Based on Traffic Identification

<p align="justify">Due to the growing variety of data center services, the bursty and variability of data traffic is increasing. In order to make the network better meet the needs of upper-layer services, it is necessary to design a more flexible optical internet topology reconstruction mechanisms to adapt the changing traffic demands. In the past research on optical internet, all topology reconstruction mechanisms are designed based on global data traffic. Although these mechanisms can fully utilize the flexibility of the data center optical interconnection network topology and adjust topology in real time according to the traffic demands, but when the traffic is presented at the regional level, this mechanism does not give optimal results. This paper proposes a topology reconstruction mechanism for data center optical interconnection network based on traffic identification for the previously proposed data center optical switching architecture—OpenScale. The simulation results show that it utilizes the flexibility of the network to save bandwidth resources and increase the wavelength connection bandwidth utilization with a little sacrifice of throughput.</p>

Terminal and broadcast reliability analysis of direct 2-D symmetric torus network.

Reliability analysis is one of the crucial issues for any scalable optical interconnection network. Torus is a highly scalable optical interconnect for data centre networks. The traditional torus network has XY routing algorithm. We have proposed a novel optimised routing algorithm. This paper focuses on the time-dependent and time-independent analysis for both terminal and broadcast reliabilities of the torus network using XY and optimised routing algorithm under various network sizes (Ntimes N where N=8, 16, 32, 64). The results are evaluated and compared considering nodes failures in MATLAB.

Design of Hybrid Plasmonic Multi-Quantum-Well Electro-Reflective Modulators Towards <100 fJ/bit Photonic Links

Realization of on-board and inter-chip optical interconnects requires a photonic data link with power consumption well below their electrical counterparts (i.e., 50 Gb/s requires 2–4 pJ/bit/channel. External reverse-biased modulators could drastically reduce this power consumption. Here we design ultralow power GaAs/AlGaAs multi quantum well electro-reflective modulators operating at 1 V for facile integration with polymer “optical bridges”, utilizing coupled quantum confined Stark effect between adjacent quantum wells and optical coupling to hybrid surface plasmon-slab modes for significantly enhanced extinction ratio and spectral bandwidth. Distinctive from conventional electro-optical or electro-absorption modulators, this new design synergistically leverages ultra-large changes in both refractive index (|Δn|∼0.05) and absorption coefficient (Δα∼104 cm−1), achieving 35-50 dB extinction ratio at 1 V reverse bias with a low insertion loss of 1–3 dB, an incident angle tolerance of ∼5°, and a spectral bandwidth of 7–10 nm. The modulator power consumption is ∼1.9 fJ/bit without the need of thermal tuning, and the RC-limited bandwidth well exceeds 100 GHz. This new modulator enables high bandwidth and ultralow power optical interconnect networks at >100 Gb/s/channel and <100 fJ/bit/channel compatible with ever-scaling CMOS technologies.

Fabrication and transmission of optical polymer waveguide backplane for high - performance computers.

In this paper, a high-speed, large-capacity and compact optical backplane architecture for high-performance computers (HPC) is proposed and designed. The MT couplers is designed without additional mirror to divert the light from vertical cavity surface emitting laser (VCSEL) array by 90°. The light is then coupled into the optical waveguide through the MT. A bidirectional 8 channels polymer optical waveguide array with low insertion loss is designed and fabricated. The waveguides are embedded in the printed circuit board. We test the performance of 8 channels for the optical waveguide backplane. In the entire optical backplane, the averaged insertion loss of optical backplane with 50 μm input fiber is 1.62 dB with 850 nm VCSEL. The misalignment loss is 0.5 dB when the misalignment between waveguide and multi-mode fiber is 8.5 μm. Then, the transmission performances of 4 channels with different speed signals is demonstrated by off-line experiment. The optical backplane network can achieve 15 G data error-free transmission in the range of [-2 dBm, -10 dBm] received optical power (ROP). Based on VCSEL arrays, the optical interconnection network system can achieve 8 channels parallel signal transmission. In the optical backplane, the 10 Gbit / s data generated / processed by field programmable gate array (Xilinx Kintex-7) chips can realize error-free transmission. In large-capacity, high-speed parallel HPC, the designed optical backplane system can facilitate the establishment of a large number of parallel transmissions.