Network Bandwidth Utilization Research Articles

Fuzzy PD-sliding mode control design for networked system with time delays

This research specifically addresses the problem of sliding mode control (SMC) in a particular group of T–S fuzzy systems that experience output disturbances and time delays using a networked control system. First, a model for a proportional–derivative sliding mode observer (SMO) is established, subsequently followed by the design of a controller based on SMO to maintain closed-loop system stability. The authors establish a novel mechanism to optimize the bandwidth utilization of the communication network by introducing a newly adjusted event-triggering parameter. After that, addressed a linear matrix inequality (LMI) problem to determine the controller gain and observer coefficients. This was done to guarantee the asymptotic stability of the closed-loop plant. Furthermore, the authors were able to directly identify the disturbances for these plants using the proposed descriptor SMO. Moreover, a parameter-based novel event-triggered scheme, along with a new resultant closed-loop system, will be employed to modify trigger frequency parameters within a unified framework. The controller gains will be determined by solving a linear matrix inequality, subject to certain conditions that ensure sufficiency. These conditions will be deduced to ensure asymptotic stability. In the final section, the authors presented examples from the Wind Energy System (WES) that illustrate the feasibility and effectiveness of our proposed methodology.

Optimizing bandwidth utilization and traffic control in ISP networks for enhanced smart agriculture.

As the demand for high-bandwidth Internet connections continues to surge, industries are exploring innovative ways to harness this connectivity, and smart agriculture stands at the forefront of this evolution. In this paper, we delve into the challenges faced by Internet Service Providers (ISPs) in efficiently managing bandwidth and traffic within their networks. We propose a synergy between two pivotal technologies, Multi-Protocol Label Switching-Traffic Engineering (MPLS-TE) and Diffserv Quality of Service (Diffserv-QoS), which have implications beyond traditional networks and resonate strongly with the realm of smart agriculture. The increasing adoption of technology in agriculture relies heavily on real-time data, remote monitoring, and automated processes. This dynamic nature requires robust and reliable high-bandwidth connections to facilitate data flow between sensors, devices, and central management systems. By optimizing bandwidth utilization through MPLS-TE and implementing traffic control mechanisms with Diffserv-QoS, ISPs can create a resilient network foundation for smart agriculture applications. The integration of MPLS-TE and Diffserv-QoS has resulted in significant enhancements in throughput and a considerable reduction in Jitter. Employment of the IPv4 header has demonstrated impressive outcomes, achieving a throughput of 5.83 Mbps and reducing Jitter to 3 msec.

FMPTCP: Achieving High Bandwidth Utilization and Low Latency in Data Center Networks

FMPTCP: Achieving High Bandwidth Utilization and Low Latency in Data Center Networks

Load-balancing routing algorithms for service congestion avoidance in LEO optical satellite networks

The low-Earth-orbit optical satellite network (LOSN) is an attractive solution to realize the end-to-end service quality of large-scale services, and it is the basis of establishing the 6G mobile network with high throughput and high reliability. However, the current LOSN is practically unable to achieve such high throughput for global service access. First, the centralized deployment of ground gateways will cause a heavy traffic load in the space segment of the LOSN, which becomes the bottleneck constraining the further growth of network throughput. Second, the high-speed movement of LEO satellites will cause continuous movement of traffic load, which will affect the scalability of routing computing. In this paper, two routing algorithms are proposed to increase the throughput in the dynamic LOSN topology. The congestion-aware load balancing (CALB) algorithm is proposed to address the inter-satellite link congestion in the dynamic LOSN topology. Then, a load-balancing routing algorithm based on satellite–ground cooperation (SGC-LB) is proposed to further reduce the impact of the network bottleneck. To evaluate the performance of the proposed routing algorithm, extensive simulations were performed on a 288-satellite Walker-Star constellation with inter-satellite links. The service blockage rate and network bandwidth utilization rate are evaluated showing the effectiveness of the proposed routing algorithm. The network using the SGC-LB algorithm can accommodate 60.00%, 56.00%, 42.22%, and 33.33% more services than using the Shortest Path algorithm, Sway algorithm, CALB algorithm, and Anycast algorithm, respectively, with zero service congestion during the simulation. The SGC-LB also gains a 6.04%, 5.00%, 5.04%, and 0.77% higher network utilization rate than the Shortest Path, Sway, CALB, and Anycast algorithms, respectively.

Lossless Data Compression for Time-Series Sensor Data Based on Dynamic Bit Packing.

In this paper, we propose a bit depth compression (BDC) technique, which performs bit packing by dynamically determining the pack size based on the pattern of the bit depth level of the sensor data, thereby maximally reducing the space wastage that may occur during the bit packing process. The proposed technique can dynamically perform bit packing according to the data's characteristics, which may have many outliers or several multidimensional variations, and therefore has a high compression ratio. Furthermore, the proposed method is a lossless compression technique, which is especially useful as training data in the field of artificial intelligence or in the predictive analysis of data science. The proposed method effectively addresses the spatial inefficiency caused by unpredictable outliers during time-series data compression. Additionally, it offers high compression efficiency, allowing for storage space savings and optimizing network bandwidth utilization while transmitting large volumes of data. In the experiment, the BDC method demonstrated an improvement in the compression ratio of up to 247%, with 30% on average, compared with other compression algorithms. In terms of energy consumption, the proposed BDC also improves data transmission using Bluetooth up to 34%, with 18% on average, compared with other compression algorithms.

Protocol-based fault detection filtering for memristive neural networks with dynamic quantization

This study examines the issue of event-triggered fault detection filtering for memristive neural networks with dynamic quantization in discrete-time domain. To facilitate digital transmissions, the system output undergoes dynamic quantized prior to transmission. Beyond the reporting event-triggered protocol, a novel event-triggered protocol is enforced, associating with dynamic quantization parameter, fault occurrence probability and network bandwidth utilization rate, to skillfully schedule the transmission frequency. A random variable that follows a binary Markov process, instead of a Bernoulli distribution, is presented to characterize the dynamic impact of denial-of-service attacks. On account of hidden Markov model and Lyapunov theory, an asynchronous filter framework is formulated to ensure stochastically stable of resulting filtering error systems. Ultimately, a simulation example is conducted to validate the usefulness of the developed methodology.

Task Offloading in Cloud-Edge Environments

Cloud computing involves transferring data to remote data centers for processing, which consumes significant network bandwidth and transmission time. Edge computing can effectively address this issue by processing tasks at edge nodes, thereby reducing the amount of data transmitted and enhancing the utilization of network bandwidth. This paper investigates intelligent task offloading under the three-layer architecture of cloud-edge-device to fully exploit the cloud-edge collaboration potential. Specifically, an optimization objective function is constructed by modelling the processing cost of all computing tasks. Additionally, asynchronous advantage actor-critic (A3C) algorithm is proposed under cloud-edge collaboration to solve the optimization problem of minimizing the sum of the weights of task offloading delay and energy consumption. Experimental results indicate that the algorithm can effectively utilize the computing resources of the cloud center, reduce task execution delay and energy consumption, and compare favourably with three existing task offloading methods.

Finite-Horizon $H_{\infty }$ Consensus Control of Discrete Time-Varying Multiagent Systems: A Dynamic Event-Based Learning Approach

This paper addresses the finite-horizon <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$H_{\infty }$</tex-math></inline-formula> consensus problem for general discrete time-varying multi-agent systems with energy-bounded external disturbances and limited network bandwidth resource. A dynamic event-based learning approach is proposed to produce a dynamic discrete-time event-triggered scheme and a dynamic distributed <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$H_{\infty }$</tex-math></inline-formula> consensus protocol. Based on the local state information, the dynamic distributed <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$H_{\infty }$</tex-math></inline-formula> consensus protocol is developed to guarantee the <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$H_{\infty }$</tex-math></inline-formula> consensus of the general discrete time-varying multi-agent systems with external disturbances. Besides, a novel criterion is established by employing the Lyapunov theory and considering the special dynamic triggering constraint, and the corresponding recursive learning algorithm is developed. In contrast with the previous works, the proposed dynamic event-based approach results in less network bandwidth utilization, and the novel recursive algorithm brings less conservatism of the designed controller. Finally, the effectiveness of the corresponding algorithm and the advantages of the proposed dynamic event-based approach are testified by a numerical simulation.

Network Bandwidth Utilization Optimization with Quantum-Inspired Approach

As the popularity of mobile devices and services continues to increase, the demand for more efficient and effective algorithms to manage network utility is increased. Here, we aim to optimize Network Bandwidth Utilization (NBU) by managing user equipments (UEs) and base stations (BSs) association in 5G ORAN architecture. We consider the network total throughput as an objective function and try to find the suitable Cell Individual Offset (CIO) to trigger UE handover with the A3 event such that the network throughput is optimal. We formulate this problem into a Quadratic Unconstrained Binary Optimization (QUBO) problem such that we can explore the power of Quantum Annealing (QA), which is a promising approach to finding the optimal solution for the NP-Hard problem. We also analyze the penalty coefficients setting problem which impacts the outcome of QA significantly. We test our QUBO formulation on a dataset of 7 BSs and 150 UEs with Fujitsu’s third generation Digital Annealer (DA3). Using our QUBO formulation with DA3 solver with fine-tuning, we can improve the throughput with 60% hit rate on this dataset.

Optimization of the Decision Criterion for Increasing the Bandwidth Utilization by Means of the Novel Effective DBA Algorithm in NG-PON2 Networks

In this paper, the reasons for the bandwidth and wavelength utilization in future next-generation passive optical networks are presented, and the possibilities for realization and utilization of extended dynamic wavelength and bandwidth algorithms for the second next-generation passive optical networks (NG-PON2) are analyzed. Next, principles of the effective dynamic bandwidth allocation are introduced in detail, focused on the importance of the decision criterion optimization. To achieve a better bandwidth utilization of dedicated wavelengths in NG-PON2 networks, this paper is focused on the novel effective dynamic bandwidth allocation algorithm with adaptive allocation of wavelengths to optical network units as well as the optimization of the decision criterion. The algorithm and the proposed method are tested and evaluated through simulation with actual traffic data. For analyzing novel extended dynamic wavelength and bandwidth algorithms used for various cases of wavelength allocation in NG-PON2 networks, the effective dynamic bandwidth allocation algorithm analysis is realized in the enhancement of simulation program. Finally, an optimization of the decision criterion defining a minimum bandwidth utilization of the actual wavelength is executed for NG-PON2 networks based on the hybrid time and wavelength division multiplexing technique.

An unambiguous TCP method for dynamic rate control to maximise bandwidth utilization in high-rate networks. (2020)

An unambiguous TCP method for dynamic rate control to maximise bandwidth utilization in high-rate networks. (2020)

Merak: An Efficient Distributed DNN Training Framework With Automated 3D Parallelism for Giant Foundation Models

Foundation models are in the process of becoming the dominant deep learning technology. Pretraining a foundation model is always time-consuming due to the large scale of both the model parameter and training dataset. Besides being computing-intensive, the pretraining process is extremely memory- and communication-intensive. These challenges make it necessary to apply 3D parallelism, which integrates data parallelism, pipeline model parallelism, and tensor model parallelism, to achieve high training efficiency. However, current 3D parallelism frameworks still encounter two issues: i) they are not transparent to model developers, requiring manual model modification to parallelize training, and ii) their utilization of computation resources, GPU memory, and network bandwidth is insufficient. We propose <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Merak</i> , an automated 3D parallelism deep learning training framework with high resource utilization. Merak automatically deploys 3D parallelism with an automatic model partitioner, which includes a graph-sharding algorithm and proxy node-based model graph. Merak also offers a non-intrusive API to scale out foundation model training with minimal code modification. In addition, we design a high-performance 3D parallel runtime engine that employs several techniques to exploit available training resources, including a shifted critical path pipeline schedule that increases computation utilization, stage-aware recomputation that makes use of idle worker memory, and sub-pipelined tensor model parallelism that overlaps communication and computation. Experiments on 64 GPUs demonstrate Merak's capability to speed up training performance over state-of-the-art 3D parallelism frameworks of models with 1.5, 2.5, 8.3, and 20 billion parameters by up to 1.42, 1.39, 1.43, and 1.61×, respectively.

Smart Utilities IoT-Based Data Collection Scheduling

The Internet of Things is an ecosystem that connects billions of smart devices, meters, and sensors. These devices and sensors collect and share data for use and evaluation by organizations in different industry sectors. Humans may use the IoT to live and work more intelligently and gain total control over their lives. Consequently, IoT can be used to connect devices and integrate them with new digital technologies for customers. On the other hand, smart utility companies in the electric, gas, and water sectors need to deliver services more efficiently and analyze their operations in a way that can help optimize performance, detect growing problems in real time, and initiate fixes to avoid unplanned service interruptions. Building actual smart metering networks is costly and time-consuming. Therefore, in this paper, a new Smart Utilities Traffic Scheduling Algorithm (SUTSA) is proposed. To minimize the system complexity, the model is based on narrowband power line communication, in which a wired hidden network sends data across power lines. A simulation is performed using OPNET Modeler 14.5 to evaluate the proposed model. The results proved that the proposed model is highly scalable and achieves full network-bandwidth utilization in different situations based on different application requirements.

Misbehavior detection system with semi-supervised federated learning

V2X communications can enhance transportation safety by exchanging safety information between vehicles, road infrastructures, networks, and pedestrians. However, the safety messages are vulnerable to disruption from faulty components or an attack that can cause misinformation. Recently, a machine learning-based misbehavior detection system (MDS) has been widely investigated to detect the misbehaving vehicles to secure the V2X communications. Nonetheless, machine learning models need sufficient labeled data for learning purposes. However, the volume of unlabeled data is usually larger than that of labeled data in practice. Moreover, transferring the large dataset to a centralized learning model will consume much bandwidth. Thus, we propose a semi-supervised federated learning MDS to overcome the limitations of unlabeled data and bring the training close to the data sources to reduce the bandwidth to the core network. Overall, our model with only limited labeled data training (5%–30%) can achieve the F1-score up to 0.96 and the recall up to 0.95. The F1-score is up to 0.26 higher and the recall is up to 0.29 higher than the performance of centralized supervised learning. The federated learning model can reduce the core network bandwidth utilization by up to 95%.

Quantized filtering for networked Takagi–Sugeno fuzzy systems with multipath data packet dropouts

This paper concerns the mixed and passivity‐based filtering problem for networked nonlinear systems with randomly occurring parameter uncertainties, multipath data packet dropouts, time‐varying delay, and quantization effects. The discrete‐time nonlinear plant considered in this paper is modeled as a Takagi–Sugeno fuzzy system with plant rules. The data missing phenomenon is considered in both measurement and performance output signals due to the unreliable nature of communication links. Further, to reduce the network bandwidth utilization, the measurement quantization is also employed for both the outputs before transmitting through the communication channel. Also, additive filter gain fluctuation is taken into account to reflect the imprecision in filter implementation. Stochastic variables obeying the Bernoulli distribution are incorporated in the model description to characterize the random occurrence of parameter uncertainties and data packet dropouts. A set of sufficient conditions assuring the stochastic stability of the filtering error system with mixed and passivity performance is derived according to Lyapunov stability theory. Finally, three numerical examples including mass–spring–damper and tunnel diode circuit model are established to establish the validity of the developed filter design algorithm.

Tricube Weighted Linear Regression and Interquartile for Cloud Infrastructural Resource Optimization

Cloud infrastructural resource optimization is the process of precisely selecting the allocating the correct resources either to a workload or application. When workload execution, accuracy, and cost are accurately stabilized in opposition to the best possible framework in real-time, efficiency is attained. In addition, every workload or application required for the framework is characteristic and these essentials change over time. But, the existing method was failed to ensure the high Quality of Service (QoS). In order to address this issue, a Tricube Weighted Linear Regression-based Inter Quartile (TWLR-IQ) for Cloud Infrastructural Resource Optimization is introduced. A Tricube Weighted Linear Regression is presented in the proposed method to estimate the resources (i.e., CPU, RAM, and network bandwidth utilization) based on the usage history in each cloud server. Then, Inter Quartile Range is applied to efficiently predict the overload hosts for ensuring a smooth migration. Experimental results show that our proposed method is better than the approach in Cloudsim under various performance metrics. The results clearly showed that the proposed method can reduce the energy consumption and provide a high level of commitment with ensuring the minimum number of Virtual Machine (VM) Migrations as compared to the state-of-the-art methods.

Machine Learning at the Mobile Edge: The Case of Dynamic Adaptive Streaming Over HTTP (DASH)

Dynamic Adaptive Streaming over HTTP (DASH) is a standard for delivering video in segments and adapting each segment’s bitrate (quality), to adjust to changing and limited network bandwidth. We study segment prefetching, informed by machine learning predictions of bitrates of client segment requests, implemented at the network edge. We formulate this client segment request prediction problem as a supervised learning problem of predicting the bitrate of a client’s next segment request, in order to prefetch it at the mobile edge, with the objective of jointly improving the video streaming experience for the users and network bandwidth utilization for the service provider. The results of extensive evaluations showed a segment request prediction accuracy of close to 90% and reduced video segment access delay with a cache hit ratio of 58%, and reduced transport network load by lowering the backhaul link utilization by 60.91%.

A Parallelizable Task Offloading Model with Trajectory-Prediction for Mobile Edge Networks

As an emerging computing model, edge computing greatly expands the collaboration capabilities of the servers. It makes full use of the available resources around the users to quickly complete the task request coming from the terminal devices. Task offloading is a common solution for improving the efficiency of task execution on edge networks. However, the peculiarities of the edge networks, especially the random access of mobile devices, brings unpredictable challenges to the task offloading in a mobile edge network. In this paper, we propose a trajectory prediction model for moving targets in edge networks without users’ historical paths which represents their habitual movement trajectory. We also put forward a mobility-aware parallelizable task offloading strategy based on a trajectory prediction model and parallel mechanisms of tasks. In our experiments, we compared the hit ratio of the prediction model, network bandwidth and task execution efficiency of the edge networks by using the EUA data set. Experimental results showed that our model is much better than random, non-position prediction parallel, non-parallel strategy-based position prediction. Where the task offloading hit rate is closed to the user’s moving speed, when the speed is less 12.96 m/s, the hit rate can reach more than 80%. Meanwhile, we we also find that the bandwidth occupancy is significantly related to the degree of task parallelism and the number of services running on servers in the network. The parallel strategy can boost network bandwidth utilization by more than eight times when compared to a non-parallel policy as the number of parallel activities grows.

Event-Triggered Fault Detection Filtering of Fuzzy-Model-Based Systems With Prescribed Performance

This article deals with the fault detection filtering problem for a nonlinear dynamic system in the Takagi–Sugeno fuzzy framework. An event-triggered scheme considering network bandwidth utilization rate and fault occurrence probability is utilized, which can conserve communication resources and reduce computational burden. The fault detection fuzzy filter is used as a residual generator, and the proposed event-based fault detection scheme is formulated as a fuzzy filtering problem. Moreover, the generated residual is robust against exogenous disturbances while being sensitive to system faults, and the resulting fault detection system has asymptotic stability with a specified <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\mathcal {H}_{\infty }$</tex-math></inline-formula> error property. The corresponding solvability conditions of the fault detection filter are constructed by converting the nonconvex feasibility problem into a series of optimization problems. Finally, a numerical simulation is conducted to demonstrate the feasibility and validity of the proposed scheme.

MiCS

Existing general purpose frameworks for gigantic model training, i.e., dense models with billions of parameters, cannot scale efficiently on cloud environment with various networking conditions due to large communication overheads. In this paper, we propose MiCS, which Minimizes the Communication Scale to bring down communication overhead. Specifically, by decreasing the number of participants in a communication collective, MiCS can utilize heterogeneous network bandwidth, reduce network traffic over slower links, reduce the latency of communications for maintaining high network bandwidth utilization, and amortize expensive global gradient synchronization overhead. Our evaluation on AWS shows that the system throughput of MiCS is up to 2.89× that of the state-of-the-art large model training systems. MiCS achieves near-linear scaling efficiency, which is up to 1.27× that of DeepSpeed. MiCS allows us to train a proprietary model with 100 billion parameters on 512 GPUs with 99.4% weak-scaling efficiency, and it is able to saturate over 54.5% theoretical computation power of each GPU on a public cloud with less GPU memory and more restricted networks than DGX-A100 clusters.