Microsecond Level Research Articles

Analog In-memory Circuit Design of Polynomial Multiplication for Lattice Cipher Acceleration Application

As the core operation of lattice cipher, large-scale polynomial multiplication is the biggest computational bottleneck in its realization process. How to quickly calculate polynomial multiplication under resource constraints has become an urgent problem to be solved in the hardware implementation of lattice ciphers. Therefore, an analog in-memory circuit for fast polynomial multiplication calculation is proposed. First, an in-memory computing circuit for Discrete Fourier Transform and Inverse Discrete Fourier Transform based on memristor array is designed. On this basis, a fully analog circuit that can realize polynomial multiplication in one step is designed. Compared with traditional hardware implementation, the in-memory calculation method used in this article decreases the calculation time of polynomial multiplication to the microsecond level, which greatly improves the speed of lattice cipher encryption and decryption. For the specific examples in this article, PSPICE simulation shows that the average accuracy of the calculation result is above 99.90%.

Theoretical MRCI + Q study on electronic structure and spectroscopic properties of the HX+(X = F, Cl, Br, I) cations

In this work, we have systematacially investigated the electronic structure and spectroscopic properties of the HX+ (X = F, Cl, Br, I) cations by using the highly correlated MRCI + Q approach. This is the first comprehensive ab initio study on the electronic states of the HX+ cations. The spin–orbit coupling (SOC) effect is introduced with the state interaction approach. There are total 10 Λ-S and 18 Ω states obtained in the calculation. The results show that the potential energy curves’ (PECs) shapes and structure of the electronic states of different HX+ cations exhibit the significant distinction due to the X-dependent energy-level order of the unique dissociation channel H+(1Sg) + X(2Pu). From PECs, the spectroscopic constants of the bound electronic states are determined, which are in good agreement with available observed values. Regarding HBr+ and HI+, the good agreement has been achieved only when the SOC effect is considered. The predissociation for the first excited A2Σ+ state of HCl+, HBr+, and HI+ is analyzed based on computed spin–orbit matrix elements. Around the equilibrium position, the energy splittings of X2Π are calculated to be 288, 661, 2691, and 5137 cm−1 for HF+, HCl+, HBr+, and HI+, respectively. It has been demonstrated that SOC is substantial for HX+, leading to significant changes in PECs’ shapes as well as in electronic structure. Finally, the transition properties are predicted, including transition dipole moments, Franck–Condon factors, and radiative lifetimes. Both the Ω transitions A2Σ+ 1/2-X2Π3/2 and A2Σ+ 1/2-X2Π1/2 of the HX+ cations are determined to possess the radiative lifetimes at the microsecond (µs) level.

Hydraulic-mechanical-electrical coupled model framework of variable-speed pumped storage system: Measurement verification and accuracy analysis

With the large-scale operation of wind and solar energy, the demand for energy storage in power grids has increased sharply. As a reliable means of long-term energy storage, the variable-speed pumped-storage power station (VSPSU) is a new development direction for pumped storage that has attracted increasing attention owing to its advantages of more rapid and flexible action. However, accurately predicting the transient process of a VSPSU is a key challenge in its design and operation. The inconsistent time step required for the fine simulation of the hydraulic-machine subsystem (millisecond level) and the electrical subsystem (microsecond level) is the main challenge that hinders the coordination of system simulation efficiency and accuracy and restricts further development of the entire coupling system simulation. Simultaneously, a coupling refinement model verified by multi-condition measurements has not been reported. This paper proposes a VSPSU hydraulic-mechanical-electrical coupling model framework that has been verified experimentally. The innovation of the model framework lies in the realization of the fine solution of the hydraulic and electrical systems through the coupling strategy of the variable simulation step size rather than using model simplification. Based on a real micro-VSPSU system, we conducted experiments under various operating conditions and compared the measurements with the numerical simulation results. The results showed that in the generation mode, the maximum simulation error of the proposed coupling model framework on the hydraulic-mechanical side was 2.29 %, and the maximum simulation error on the electrical side was 9.59 %. Finally, we quantified the contribution of key parameters in the model framework to the dynamic response process and clarified the influence of physical parameters on the transient process in the VSPSU. Compared with conventional models, the advantage of the proposed coupling framework is that it can realize the coupling of a hydraulic-mechanical-electrical system without simplification, which has been verified by measuring multiple working conditions. This framework provides a reliable tool for the parameter design and operation optimization of VSPSU.

Research on FPGA-based dual-channel time synchronization system

As geological exploration progresses in depth and scope, the distance between transmitting and receiving systems also increases. Consequently, the synchronization of time between the two becomes an increasingly critical issue. In response, this paper presents a proposed design for a time synchronization system that is suitable for use between transmitting and receiving equipment. The system uses GPS as the main clock source. The information within the GPS system will be encoded and decoded using the FPGA. This allows us to obtain both UTC time and the time of the IRIG-B code format, enabling compatibility with the various system interfaces. Our experiments have confirmed that the system has a time error at the microsecond level, and the precision satisfies the design requirements. This verifies the validity and reliability of the system.

Efficient charge separation and improved photoelectric properties of GaN/WS2/MoS2 heterojunction: A molecular dynamics simulations

Utilizing the stacking of two-dimensional (2D) materials to fabricate heterojunction represents an exceedingly effective approach for designing high-performance optoelectronic devices. In this paper, the GaN/WS2/MoS2 heterojunction was systematically studied using DFT calculations. The interface effect caused the effective separation of photo-excited carriers in the GaN/WS2/MoS2 heterojunction by inducing a stepwise transfer of the carriers, forming a build-in electric field. Due to the decrease in bandgap, the heterojunction exhibited enhanced light absorption ability across the entire spectral region. Especially, the carrier recombination time reached the microsecond level, greatly extending the lifetime of photo-excited carriers and effectively improving the energy conversion performance of photodetectors. Moreover, the augmentation of tensile strain enhances both the self-powering capacity and the near-infrared light detection capability of the GaN/WS2/MoS2 heterojunction. The exceptional performances exhibited by the GaN/WS2/MoS2 heterojunction make it a highly up-and-coming candidate for optoelectronics application.

Directional fracturing of granite cylinder via nitromethane-based energetic materials driven by electrical wire explosion

The combination of electrical wire explosion and low sensitivity energetic materials is an efficient and practical method to generate controllable strong pressure waves, which is a potential alternative to explosives in directional fracturing applications. In this paper, a new configuration of hybrid load is proposed to generate synchronized pressure waves for directional fracturing, with the advantages of lower energy storage and smaller blast-hole diameters. The pressure wave of hybrid load explosion is measured, and the explosion synchronicity of the two loads at the microsecond level is verified by self-illumination images. Then, two-hole granite cylinders are used to verify the feasibility of directional fracturing, and the results show that the double loads explosion formed directional fractures connecting the two holes and a horizontal failure plane. The fracturing process, including fracture initiation, propagation, and connection, is recorded with a high-speed camera and is reproduced in numerical simulation. Based on the experimental and numerical results, the fracture morphology is characterized as four areas, and the fracturing mechanism is analyzed.

An 82-nW 0.53-pJ/SOP Clock-Free Spiking Neural Network With 40-μs Latency for AIoT Wake-Up Functions Using a Multilevel-Event-Driven Bionic Architecture and Computing-in-Memory Technique

This article presents a clock-free spiking neural network (SNN) intelligent inference engine (IIE) for artificial intelligence of things (AIoT) sensor nodes, which often operate in random-sparse-event (RSE) scenarios. The IIE drastically reduces the system’s long-term average (LTA) power consumption, improves energy efficiency, and achieves microsecond level inference latency. Three techniques are proposed: 1) A clock-free SNN architecture without clock tree, frame generator, and arbiter, is driven by the output spikes, which are encoded with level-crossing (LC) sampling method; the circuit activity is completely related to event activity and spike rates, dramatically reducing the overall power consumption and latency. 2) The bioinspired leaky-integrate-fire (LIF) neurons directly extract the time-domain information from asynchronous spikes, reducing the network size and number of operations. 3) The computing-in-memory (CIM) and mixed-signal synapse-neuron circuits are employed to increase the SNN parallelism and avoid weight movements, thus improving the energy efficiency and response speed. The measured LTA power is bounded at 82 nW while the event-driven chip is on call and waiting for events; the energy efficiency is 0.53 pJ per synapse operation (SOP), only 1/3 that of state-of-the-art methods at 4bit weights even with 180 nm technology. We demonstrate electrocardiogram (ECG) recognition as a typical AIoT application, and the power consumption is less than 350 nW. The measured accuracy of abnormal ECG detection is 90.5%. Moreover, the latency is only 40 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\mu $</tex-math> </inline-formula> s to realize real-time NN inference. This work provides an effective solution for AIoT nodes that require both ultralow power and fast response.

Tissue Ablation Using Irreversible Electrolytic Electroporation with Reduced Voltage

Thermal tissue ablation may damage surrounding healthy tissue and cause pain. In this study, tissue ablation with the sequential application of electrical energy-inducing irreversible electroporation (IRE) and electrolysis (EL) (IRE + EL = IREEL) was investigated. An IREEL device was designed to control five output pulse parameters: voltage level (VL), pulse width (PW), pulse interval (PI), pulse number (PN), and pulse tail time (PTT). IREEL experiments were conducted on vegetable tissue. The results indicated that by increasing the VL and PTT, the ablation area increased, whereas the impedance was reduced significantly. Almost no ablation area was observed when only EL or IRE at 500 V and 1000 V, respectively, were applied. The ablation area observed with IRE alone at 1500 V was defined as 100%. In the case of IREEL at 500 V and 1000 V, ablation was induced even with the use of micro-second level PTT, and ablation areas of 91% and 186% were achieved, respectively. For IREEL at a voltage of 1500 V, the ablation area expanded to 209% and the maximum temperature was 48.7 °C, whereas the temperature did not exceed 30 °C under other conditions. A change in pH was also observed in an agar-gel phantom experiment which was conducted to examine and confirm whether IREEL induced electrolysis. IREEL was able induce ablation at low voltages owing to the synergistic effect of applying IRE and EL sequentially. Moreover, the ablation areas at high voltages could be increased compared to the areas observed when IRE and EL were applied independently.

Assessing the Current State of Amber Force Field Modifications for DNA─2023 Edition.

Advances in molecular dynamics (MD) software alongside enhanced computational power and hardware have allowed for MD simulations to significantly expand our knowledge of biomolecular structure, dynamics, and interactions. Furthermore, it has allowed for the extension of conformational sampling times from nanoseconds to the microsecond level and beyond. This has not only made convergence of conformational ensembles through comprehensive sampling possible but consequently exposed deficiencies and allowed the community to overcome limitations in the available force fields. The reproducibility and accuracy of the force fields are imperative in order to produce biologically relevant data. The Amber nucleic acid force fields have been used widely since the mid-1980s, and improvement of these force fields has been a community effort with several artifacts revealed, corrected, and reevaluated by various research groups. Here, we focus on the Amber force fields for use with double-stranded DNA and present the assessment of two recently developed force field parameter sets (OL21 and Tumuc1). Extensive MD simulations were performed with six test systems and two different water models. We observe the improvement of OL21 and Tumuc1 compared to previous generations of the Amber DNA force. We did not detect any significant improvement in the performance of Tumuc1 compared to OL21 despite the reparameterization of bonded force field terms in the former; however, we did note discrepancies in Tumuc1 when modeling Z-DNA sequences.

SiC MOSFET Trap Characterization Based on the Self-built Test Platform

The interface state of SiC/SiO2 is one of the key factors limiting the reliability and performance of SiC MOSFETs. In this paper, we built a dedicated detrap measurement platform to improve the accuracy of trap measurement. Based on the excellent voltage switching time of this platform, the entire voltage switching process can be controlled within 1 µs, and the time resolution is also improved. The original millisecond-level acquisition accuracy is improved to a microsecond level, and the identification range of traps is expanded. The traps were extracted by using the transient current method based on the Bayesian deconvolution algorithm. With this approach, we investigated the trapping mechanism of SiC MOSFETs and mainly characterized the trap locations, energy levels, and trapping time constants. The findings revealed the existence of three different types of traps/defects, Dp1, Dp2, and Dp3, with activation energies of 0.28 eV, 0.035 eV, and 0.084 eV, respectively. The non-destructive characterization of SiC MOSFET defects can be realized through the test platform and Bayesian deconvolution algorithm in this paper, which brings great convenience for trap characterization.

A Real-Time Maximum Efficiency Tracking for Wireless Power Transfer Systems Based on Harmonic-Informatization

With the increased requirements of dynamic wireless charging scenarios, the industry puts forward higher technical specifications for system efficiency. However, the traditional millisecond level response time cannot meet the requirements of dynamic maximum efficiency tracking (MET) system for real-time control. Therefore, this article proposes a wireless power transfer system based on harmonic-informatization, which simplifies the circuit through the harmonic and bandstop filter, so that the mutual inductance can be extracted by harmonic, and then MET can be implemented in real time. Compared with the traditional fundamental system, the harmonic-informatization system has the advantages of no iteration, no communication, and simple calculation. Because the operation of MET under harmonic is simple, the experimental system in this article successfully achieves the microsecond level response time. At the same time, the experiment shows that the system efficiency and power change little after adding the filter, which is due to the low equivalent reactance of the filter under the fundamental.

Regression-decision-tree based parameter optimization of measurement-device-independent quantum key distribution

The parameter configuration of quantum key distribution (QKD) has a great effect on the communication effect, and in the practical application of the QKD network in the future, it is necessary to quickly realize the parameter configuration optimization of the asymmetric channel measurement-device-independent QKD according to the communication state, so as to ensure the good communication effect of the mobile users, which is an inevitable requirement for real-time quantum communication. Aiming at the problem that the traditional QKD parameter optimization configuration scheme cannot guarantee real-time, in this paper we propose to apply the supervised machine learning algorithm to the QKD parameter optimization configuration, and predict the optimal parameters of TF-QKD and MDI-QKD under different conditions through the machine learning model. First, we delineate the range of system parameters and evenly spaced (linear or logarithmic) values through experimental experience, and then use the traditional local search algorithm (LSA) to obtain the optimal parameters and take them as the optimal parameters in this work. Finally, we train various machine learning models based on the above data and compare their performances. We compare the supervised regression learning models such as neural network, K-nearest neighbors, random forest, gradient tree boosting and classification and regression tree (CART), and the results show that the CART decision tree model has the best performance in the regression evaluation index, and the average value of the key rate (of the prediction parameters) and the optimal key rate ratio is about 0.995, which can meet the communication needs in the actual environment. At the same time, the CART decision tree model shows good environmental robustness in the residual analysis of asymmetric QKD protocol. In addition, compared with the traditional scheme, the new scheme based on CART decision tree greatly improves the real-time performance of computing, shortening the single prediction time of the optimal parameters of different environments to the microsecond level, which well meets the real-time communication needs of the communicator in the movable state. This work mainly focuses on the parameter optimization of discrete variable QKD (DV-QKD). In recent years, the continuous variable QKD (CV-QKD) has developed also rapidly. At the end of the paper, we briefly introduce academic attempts of applying machine learning to the parameter optimization of CV-QKD system, and discuss the applicability of the scheme in CV-QKD system.

BEATS: An Open-Source, High-Precision, Multi-Channel EEG Acquisition Tool System.

Stable and accurate electroencephalogram (EEG) signal acquisition is fundamental in non-invasive brain-computer interface (BCI) technology. Commonly used EEG acquisition systems' hardware and software are usually closed-source. Its inability to flexible expansion and secondary development is a major obstacle to real-time BCI research. This paper presents the Beijing University of Posts and Telecommunications EEG Acquisition Tool System named BEATS. It implements a comprehensive system from hardware to software, composed of the analog front end, microprocessor, and software platform. BEATS is capable of collecting 32-channel EEG signals at a guaranteed sampling rate of 4kHz with wireless transmission. Compared to state-of-the-art systems used in many EEG fields, it displays a better sampling rate. Using techniques including direct memory access, first in first out, and timer, the precision and stability of the acquisition are ensured at the microsecond level. An evaluation is conducted during 24 hours of continuous acquisitions. There are no packet losses and the average maximum delay is only 0.07s/h. Moreover, as an open-source system, BEATS provides detailed design files, and adopts a plug-in structure and easy-to-access materials, which makes it can be quickly reproduced. Schematics, source code, and other materials of BEATS are available at https://github.com/buptantEEG/BEATS.

High-performance self-driven near-infrared photodetector based on single-walled carbon nanotube/graphene/Al2O3/InP

Carbon nanotube (CNT) is one of the typical materials for near-infrared photodetectors. Combining its ultra-high mobility, excellent infrared absorption, and suitable energy band structure with graphene and semiconductors, it greatly improves the photodetector performance. This paper reports a self-driven near-infrared photodetector of single-walled carbon nanotubes (SWCNTs)/single-layer graphene/Al 2 O 3 /P-type InP structure, the photocurrent of the device is greatly improved with the help of SWCNTs, and the Al 2 O 3 layer improves the Schottky barrier between InP and graphene, which can suppress the carrier recombination on the semiconductor surface. This device structure exhibits a highly sensitive photoresponse at 808 nm, and at zero under the bias condition, the responsivity reaches 154.3 mA·W -1 , and the detection rate reaches 1.27 × 10 12 Jones. It has a fast response time of microsecond level, and can meet the detection requirements in the mid-frequency range. This result is expected to provide ideas for further optimization on single-walled carbon nanotubes in near-infrared photodetectors. • A monolayer graphene/InP Schottky junction near-infrared photodetector with SWCNTs is fabricated. • The combination of SWCNTs and graphene greatly improves the responsivity, detectivity, and quantum efficiency of the device. • The interface passivation technology of Al 2 O 3 effectively reduces dark current. • The ballistic transport mechanism of SWCNTs enables the device with microsecond response time.

Impacts of microsecond control in laser powder bed fusion processing

Lasers used in standard metal Laser Powder Bed Fusion (LPBF) machines are focused at the build plane with typical beam diameters less than 100 µm. This results in millions of individual laser beam scanning vectors when fabricating complex metal parts, and in most build scenarios, the laser turns on and off at the beginning and end of each vector, respectively. The standard configuration for LPBF machines includes a laser beam steering system with two mirrors attached to galvanometers, called the scanner. Most commercial lasers have response times for beam power control below 10 µs yet acceleration/deceleration times of commercial scanners are an order of magnitude slower (due to the dynamics of the mechanical system). A common strategy for allowing the scanner to “catch up” with the commanded location and speed is to employ “skywriting” with additional built-in time delays to allow for the scanner to maintain constant velocity while the laser is on. These delays along with skywriting are intended to increase resolution and reduce defects, although the actions and times are typically not accessible by the user. In fact, the millions of laser on/off events result in additional sources for defect generation to include lack of fusion voids, keyhole voids, surface roughness, and spatter. Ultimately, this results in differences in part performance within builds (build location) and between LPBF platforms. Recent work has identified methods to eliminate some forms of laser on/off defects by controlling the laser power ramp rate. To the authors knowledge, laser power ramping is not currently implemented on any commercial LPBF machine, and the use of laser power ramping has yet to be evaluated on realistic, complex parts. As a result, the current work implemented laser power ramping on a commercial LPBF machine using a complicated geometry with 1,309,338 laser on/off events (vectors) to analyze the holistic impact of fabricating real parts on defects, mechanical properties, and spatter. Ramping the laser power over 300µs eliminated the formation of back spatter, and a demonstration build of a complex qualification test artifact using IN718 reduced total spatter by nearly 5%. However, this laser power ramp change over 300µs (i.e., linearly varying laser power from 0 W to the desired power set point over 300µs) increased lack of fusion defects near part edges that reduced the tensile ductility from 17% to 12% without affecting tensile strength. The research highlighted the financial challenges imposed when scrutinizing microsecond level interactions, and as a result, a low-cost method to identify minimum ramp rates to prevent back spatter using a smart phone and an optical microscope was demonstrated. Microsecond control over the laser must be accompanied by microsecond control of the scanner and the impacts must be quantified holistically over a variety of response variables to broadly apply the developments to industry applications.1

A Survey of Storage Systems in the RDMA Era

Remote Direct Memory Access (RDMA) based network devices are increasingly being deployed in modern data centers. RDMA brings significant performance improvements over traditional network devices such as Ethernet due to its unique features: <i>protocol offloading</i> and <i>memory semantics</i> . In particular, it can achieve microsecond level latency, which is about 2 <inline-formula><tex-math notation="LaTeX">$\sim$</tex-math></inline-formula> 3 orders of magnitude improvement. With such improvement in hardware, the software stack, including device drivers and programming libraries, is becoming a new performance bottleneck. Developers need to use new programming libraries to take full advantage of the performance of the underlying hardware. Storage systems are very important in modern data centers. This article surveys the current efforts to use RDMA for optimizing storage systems. We first present five classes of RDMA-based storage systems, including key-value stores, file systems, distributed memory systems, database systems, and systems using smart NICs, to demonstrate different design choices. Then, we examine the core modules of storage systems from different perspectives: communication mode, concurrency control, fault tolerance, caching, and resource management. Finally, we provide some design guidelines for new RDMA-based storage systems, as well as a discussion of opportunities and challenges.

A Test Method for Response Behavior of Metal-Oxide Arrester Subjected to Transient Electromagnetic Disturbances

Metal-oxide arresters (MOA) are widely used to protect equipment against transient electromagnetic disturbances (TED) like very fast transient overvoltage (VFTO), high-altitude electromagnetic pulse (HEMP), lightning pulse, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">et al.</i> Researches on the performance of MOA under microsecond level rise-time TED are abundant, such as lightning pulse. However, when MOA is subjected to TED of rise-time at nanosecond level, the voltage/current responses of MOA are not fully studied due to the lack of a standardized and reliable test setup. In this paper, a test setup is proposed for acquiring the response behavior of MOA under nanosecond level pulse. The setup is comprised of a pulse generator, a two-branch transmission line system and a high voltage power divider. The generator can provide stable double exponential pulse with tunable rise-time ranging from 5 ns to 100 ns. The divider ensures an even distribution of voltage/current surge wave along the two-branch line. Through the numerical analysis, both impinging and residual surge waveforms could be retrieved from measurements. The proposed method has been applied to investigate response behavior of MOA subjected to nanosecond level rise-time pulses.

Preparation and operating characteristics analysis of high-speed iron phthalocyanine organic phototriode

The structure of Cu/FePc/Al/FePc/Cu vertical structure organic bipolar junction triode (VOBJT) is prepared by vacuum vapour deposition and magnetron sputtering process, using the photosensitive organic semiconductor material iron phthalocyanine (FePc) as the organic layer. The experiment results show that the device’s output current is unsaturated, and the base voltage can effectively regulate the output current. FePc-VOBJT has good optoelectronic characteristics, and the optical response rate can reach R = 0.19 A W−1. Under the AC base voltage, the cut-off frequency of FePc-VOBJT is f c = 10.7 kHz, and the turn-on and turn-off time can reach the microsecond level. When FePc-VOBJT responds to light pulse, the cut-off frequency f c(Light) = 2 kHz, and the switch time can reach the millisecond level. Vertical structure dramatically reduces the device’s conductive channel length and junction capacitance, which makes the organic semiconductor materials with poor electrical properties expected to be used in high-speed photoelectric sensors.

Experimental investigations of the effect of millisecond-delay time on the blast vibration reduction with electronic detonators

The error of precise initiation time controlled at the microsecond level in the electronic detonator blasting is accurate enough to realize blasting vibration reduction; however, the realization of precise control of blasting vibration using the electronic detonators remains rudimentary, researches on the precise zoning protection of buildings and field experiments are lacked. Therefore, multi-delay time of double-hole and multi-hole blasting experiments were carried out to explore the influence of different delay time on blast-induced PPV (Peak Particle Velocity). Based on the measured blasting vibration waveforms with blasting center distance of 16–200 m the reasonable delay time reducing PPV was investigated when the protective buildings at different blasting center distances and the reasonable delay time suitable for zoning protection was put forward to verify the applicability of the electronic detonators. The results show that as the initiation delay time increases, the PPV of blasting vibration first decreases then increases and finally attenuates to the single-hole PPV, with the increase of the blasting center distance, the reasonable delay time for reducing PPV gradually increases. However, a greater vibration enhancement effect will be produced when the delay time reaches at a certain value in the near area, the generated maximum PPV is about 1.8 times the single-hole PPV and there will be multiple short delay time producing enhancement effect in multi-hole blasting, the short delay time cannot be blindly adopted to reduce vibration. Therefore, the delay time of electronic detonators should be reasonably set to achieve the effect of blasting vibration reduction based on the distance between the protected buildings and the blasting zone.

Dynamic Distributed Multi-Path Aided Load Balancing for Optical Data Center Networks

Benefiting from dense connections in data center networks (DCNs), load balancing algorithms are capable of steering traffic into multiple paths for the sake of preventing traffic congestion. However, given each path&#x2019;s time-varying and asymmetrical traffic state, this may also lead to worse congestion when some paths are overutilised. Especially in the two-tier hybrid optical/electrical DCNs (Hoe-DCNs), the port contentions and large-grained optical packets of the fast optical switch (FOS) require the top-of-rack (TOR) switch to have microsecond-level load balancing capability for microburst traffic. This paper establishes a leaf-spine Hoe-DCN model to illustrate the principal characteristic of dynamic load balancing in TOR switches for the first time. Moreover, we propose the dynamic distributed multi-path (DDMP) load balancing algorithm that relies on dynamic hashing computing for network flow distribution in DCNs, which dynamically adjusts traffic flow distribution at microsecond level according to the inverse ratio of the buffer occupancy. The simulation results show that our proposed algorithm reduces the TOR-to-TOR latency by 15.88&#x0025; and decreases the packet loss by 22.06&#x0025; compared to conventional algorithms under regular load conditions, which effectively improves the overall performance of the Hoe-DCNs. Moreover, our proposed algorithm prevents more than 90&#x0025; packet loss under low load conditions.