Search

For supporting the upcoming applications, wireless mobile users require high bandwidth. To serve these requirements, a wireless network that has an optical backbone with a colossal bandwidth capacity is required. The radio-over-fiber (RoF) network is a promising solution that can provide high bandwidth. To accomplish this goal, we propose a converged medium access control (MAC) protocol for the RoF network, which uses a single-gate or multi-gate polling algorithm for time and wavelength allocation to the users. We propose an analytical delay model for single and multiple wavelength architectures, considering the user identification and data transfer delay for single-gate and multi-gate polling algorithms. We verify it with the results obtained from simulations. Finally, we present the mean delay and packet loss rate of the proposed algorithm by varying the user load and traffic burstiness and prove that the proposed algorithm gives better delay performance and a lower packet loss than the state-of-the-art algorithms.

The impact of the traffic characteristics of secondary users (SUs) on the performance of cognitive radio networks (CRNs) should be understood for designing operation rules. This paper focuses on multi-type burst services and network congestion problem in CRNs and evaluates the QoS of SUs based on multiple cross-layer considerations. A two-state Markov-modulate Bernoulli process (MMBP-2) is adopted to model packet flows of SUs with different burst degrees in CRNs. We propose a two-dimensional discrete queuing model to consider spectrum access, burstiness of traffic, network congestion, channel environment, user activity and finite buffer. We construct an iterative algorithm to compute the steady-state distribution of the proposed queuing model and determine the performance metrics like the throughput, the average delay, the average queue length and the total packet loss probability. Numerical analysis evaluates the QoS of SUs under different burst environments.

In recent years, the number of CPU cores in a multi-core processor keeps increasing. To leverage the increasing hardware resource, programmers need to develop parallelized software programs. One promising approach to parallelizing high-performance applications is pipeline parallelism, which divides a task into a serial of subtasks and then maps these subtasks to a group of CPU cores, making the communication scheme between the subtasks running on different cores a critical component for the parallelized programs. One widely-used implementation of the communication scheme is software-based, lock-free first-in-first-out queues that move data between different subtasks. The primary design goal of the prior lock-free queues was higher throughput, such that the technique of batching data was heavily used in their enqueue and dequeue operations. Unfortunately, a lock-free queue with batching heavily depends on the assumption that data arrive at a constant rate, and the queue is in an equilibrium state. Experimentally we found that the equilibrium state of a queue rarely happens in real, high-performance use cases (e.g., 10Gbps+ network applications) because data arriving rate fluctuates sharply. As a result, existing queues suffer from performance degradation when used in real applications on multi-core processors. In this paper, we present EQueue, a lock-free queue to handle this robustness issue in existing queues. EQueue is lock-free, efficient, and robust. EQueue can adaptively (1) shrink its queue size when data arriving rate is low to keep its memory footprint small to utilize CPU cache better, and (2) enlarge its queue size to avoid overflow when data arriving rate is in burstiness. Experimental results show that when used in high-performance applications, EQueue can always perform an enqueue/dequeue operation in less than 50 CPU cycles, which outperforms FastForward and MCRingBuffer, two state-of-the-art queues, by factors 3 and 2, respectively.

A Virtual Channel (VC) is a Time Division Multiplexed (TDM) slice of a physical channel/link. A crucial step of interconnect synthesis is to assign VCs to traffic that avoids deadlocks while meeting Power, Performance and Area (PPA) objectives. For a Network on Chip (NoC), VC assignment is tightly coupled with topology generation and routing. Inefficient VC assignment can lead to NoCs which may be an order of magnitude inferior in terms of PPA. However, both VC assignment and topology generation are combinatorially hard problems. Thus, combining VC assignment with topology generation makes it difficult to efficiently solve either.In this paper, we present a topology agnostic VC assignment approach for statically routed NoCs. This segregation enables us to solve the VC assignment problem first and then subsequently generate topology. By solving VC assignment first, we reduce the problem space of topology generation leading to NoCs with tighter PPA. We model the VC assignment problem as a Traffic Conflict Graph (TCG), capturing a global view of Quality of Service (QoS), Head-of-Line (HoL) conflicts, burstiness and external protocol level dependencies (e.g. PCIe root-complex). We apply combinatorial optimization techniques on TCG to arrive at an efficient VC assignment, that ensures deadlock free designs. The algorithm has been implemented in production NoC Synthesis tool. Results obtained on more than a dozen multi-million gate System-on-Chip (SoC) demonstrate an average improvement of 30% across metrics such as routers, resizers, power/clock domain converters, latencies etc. while meeting the performance vis-a-vis hand-tuned NoCs. The hand-tuned designs have been manually optimized over several man-months by expert designers, while the tool achieves better results in a few minutes.

The software switches and routers are rapidly developed in the NFV/SDN area, and are expected to be adopted in the emerging future mobile network (beyond-5G). Despite the high expectation of software routers, its latency characteristics have not been investigated extensively. To provide the reference latency value for software-based network function, we analyzed the latency of SRv6 Mobile User Plane on VPP software router. We show that, 1) for the beyond-5G mobile network application, the latency is in the feasible range (roughly 3–30 μs), 2) the throughput is sufficient with the range in 1–5 Mpps per CPU core, 3) the latency has the strong correlation with the input traffic load, 4) and the latency can be stable in a carefully crafted implementation. The packet interval analysis shows the burstiness of its output traffic that is enforced by the processing of a vector (bulk/batch) of packets. Our latency characteristic analysis can be of help for improving the performance of software router using the batch processing and designing software-based beyond-5G mobile communication systems.

Network traffic can reflect the operating status and resource bottleneck of the entire network. Accurate prediction of the future network is helpful in network maintenance, network optimization, routing policy design, load balancing, protocol design, and anomaly detection. However, the self-similarity, periodicity, chaos, multi-scale, and other characteristics of modern network traffic make it challenging to predict network behaviors. The available prediction models focus only on self-similarity and burstiness, lacking a more accurate and comprehensive description of the characteristics of network traffic. In this paper, we propose a prediction model based on Empirical Mode Decomposition (EMD) and the Gated Recurrent Unit (GRU) neural network with data reconstruction. First, the traffic data are reconstructed by complementing missing-points and eliminating outliers. Then, we decompose the reconstructed traffic data into several components through EMD and use each component to train the corresponding GRU neural network. Finally, the predicted values of all components are combined to get the final result. Numerical results show that the proposed prediction model offers higher accuracy and more stable performance than the state-of-the-art models.

The unpredictable variability of Data Stream Processing (DSP) application workloads calls for advanced mechanisms and policies for elastically scaling the processing capacity of DSP operators. Whilst many different approaches have been used to devise policies, most of the solutions have focused on data arrival rate and operator resource utilization as key metrics for auto-scaling. We here show that, under burstiness in the data flows, overly simple characterizations of the input stream can yet lead to very inaccurate performance estimations that affect such policies, resulting in sub-optimal resource allocation.We then present MEAD, a vertical auto-scaling solution that relies on online state-based representation of burstiness to drive resource allocation. We use in particular Markovian Arrival Processes (MAPs), which are composable with analytical queueing models, allowing us to efficiently predict performance at run-time under burstiness. We integrate MEAD in Apache Flink, and evaluate its benefits over simpler yet popular auto-scaling solutions, using both synthetic and real-world workloads. Differently from existing approaches, MEAD satisfies response time requirements under burstiness, while saving up to 50% CPU resources with respect to a static allocation.

Firewalls are located at the front line of the network against outside threats. Performance modeling and analysis of network firewalls help to better understand their behavior and characteristics. Moreover, having an analytical model in hand helps firewall designers avoid developing multiple design alternatives and thus considerably reduce the design costs. Moreover, the network administrators can proactively identify the performance bottlenecks of the network and fix them before any malicious attack which targets the network or the firewall itself. In this paper, we propose a novel analytical approach for performance modeling and analysis of network firewalls based on a discrete-time queuing system in which the bursty nature of the incoming traffic is taken into account, where traditional queuing models such as $M/M/1$ model fails to capture peculiar characteristics of the Internet traffic. Throughput, packet loss, delay, and firewalls CPU utilization are employed as performance evaluation indicators in our proposed model. In addition, we introduce a potential DoS attack with a very low rate which can be launched against firewalls with different burstiness factors.

Spectrum prediction is an important technology for spectrum cognition which constitutes the premise of cognitive radio technology. However, most of the existing spectrum prediction methods were proposed for common frequency bands. Based on this philosophy, these methods generally build a mapping model between historical statistical information and future state, and performance of the single model was verified to be limited due to that it cannot match the high burstiness of spectrum. This paper considers the prediction of bursty bands. The previous collected data of 2.4GHz Industrial, Scientific, Medical (ISM) bands is utilized for both feature analysis and method research. Specifically, the deep-reinforcement learning (DRL) method is adopted to address the high-dimensional state and action spaces caused by the classification process using the values of several features. Besides, this paper also employs the upper-limit results of the prediction to refine the classification process, i.e. those groups with better prediction performance using predictors corresponding to other classes will be readjusted. We thus name the proposed method as classification-based deep reinforcement learning (C- DRL). Finally, extensive experiments with the collected real data are conducted to verify the performance of C-DRL. The results suggest that C-DRL significantly outperforms the state-of-the-art algorithms in terms of the prediction performance.

The traffic from the Internet of Thing (IoT) characterizes self-similarity and burstiness. For time-sensitive IoT devices, how to guarantee real-time traffic forwarding is still a challenging problem. We introduce a network slicing scheme to separate a Network Slice (NS) for time-sensitive application and achieve real-time forwarding of time-sensitive traffic through the proposed buffer scheduling scheme. Departing from the conventional buffer management scheme buffering all time-sensitive traffic into the same queue, the proposed scheme adopts multi queues to buffer the packets with different latency requirements. We simulate a sliced-network 50G TDM-PON to distinguish the time-sensitive traffic and regular traffic and compare the proposed scheme with the traditional scheme. The proposed algorithm outperforms the service-loss ratio by 83.37% (load of 0.8) and reduces packet loss by 50.33% at the load of 0.3.

In this paper, we study the energy consumption of Narrowband IoT devices. The paper suggests that key to saving energy for NB-IoT devices is the usage of full Discontinuous Reception (DRX), including the use of connected-mode DRX (cDRX): In some cases, cDRX reduced the energy consumption over a 10-year period with as much as 50%. However, the paper also suggests that tunable parameters, such as the inactivity timer, do have a significant impact. On the basis of our findings, guidelines are provided on how to tune the NB-IoT device so that it meets the target of the 3GPP, i.e., a 5-Wh battery should last for at least 10 years. It is further evident from our results that the energy consumption is largely dependent on the intensity and burstiness of the traffic, and thus could be significantly reduced if data is sent in bursts with less intensity, irrespective of cDRX support.

LoRaWAN is one of the most promising standards for IoT applications. Slotted ALOHA has been recently proposed as LoRaWAnMac protocol, in an effort to achieve a performance improvement. However, the effect of traffic burstiness in a LoRaWAnenvironment has not been studied. Critical performance metrics, such as packet delay and network throughput, under bursty traffic remain to be evaluated. This paper explores the effect of bursty traffic in a Slotted-ALOHA based LoRaWAnnetwork. The network performance is studied via extensive simulation results, under traffic of different grades of burstiness (from low to high mean burst length) by using throughput and mean packet delay, as performance metrics and major conclusions are extracted.

Many real-world network data can be modelled as a bipartite graph, such as client-server in the computer network and trader-stock in the stock market. Detecting anomalies on these graphs benefit many applications, e.g., network intrusion detection and illegal trading detection. In real-world networks, these bipartite graphs are not static. Their vertices may change from time to time, and a bipartite graph may demonstrate unexpected burstiness in the form of intermittent increases or decreases in activities between vertices. In this paper, we propose a framework named BEA, a general anomaly detection framework that can effectively detect anomalies on a dynamic bipartite graph with burstiness. We evaluate our proposed framework on three public datasets against the state-of-the-art baselines. We also conduct a case study on the PayPal CSP log network. Through experiments, we show BEA's practicability and effectiveness.

The IEEE 802.1Qcr Asynchronous Traffic Shaping (ATS) proposed by Time-Sensitive Network (TSN) is a flow control mechanism based on Urgency-based Scheduler (UBS). The ATS aims to reshape flows or traffic per-hop and is still with the update of the draft.In the draft of version 2.0, the ATS algorithm concerns about two parameters cir (Committed Information Rate) and cbs (Committed Burst Size). When an ATS shaper operates in the condition of parameter consistency, i.e. the cir and the cbs have the same values of the ingress flow’s rate and burstiness respectively, the burst cascade can be avoided perfectly.On the contrary, we introduced a deterministic Network Calculus (NC) based method to compute the delay bound for an ATS hop with parameter inconsistency, with a concatenation model of this ATS shaper and its upstream Strict Priority Queuing (SPQ) scheduler. Furthermore, we induced a threshold for the cbs parameter corresponding to a certain ingress busrtiness. When the cbs of ATS is not equal to the threshold value, the delay bound of the flows shaped by the shaper will increase. In addition, the results of analysis and case study show that when the cbs of the ATS is changed, the delay bounds of other multiplexing flows will also be affected.

The rate of discovery of vulnerabilities keeps increasing, creating a problem for first responders who need to triage vulnerabilities quickly to decide where to focus their defensive efforts. One of the bottlenecks in this triaging process is the assessment of severity of vulnerabilities and the assignment of the Common Vulnerability Scoring System (CVSS) scores.In this work we study the statistical properties of the vulnerability disclosure process and make two important observations. First, we find that the time series of the number of vulnerability disclosures exhibits a long range dependence, meaning that strong correlations persist over long time periods. Such time series have high variation, high burstiness and slow convergence towards conventional estimators, such as the mean.Our second observation is that the burstiness of the vulnerability disclosure process causes delays in the analysis of vulnerabilities and as a result triaging over 40% of the vulnerabilities takes longer than the median exploit time. Hence, by the time they are analysed and assigned a CVSS score, many vulnerabilities are already being exploited.We propose techniques for modelling and analysing the vulnerability disclosure time series. We further propose reversing the order of triaging vulnerabilities and show, via simulation, that this significantly increases timely triaging of vulnerabilities, reducing the percentage of delayed assessments to 4%.

In massive and densely deployed 5G NB-IoT networks, centralized managements are computationally prohibitive and limit the network scalability. As a result, the auto-integration and self-management capabilities are required in the 5G NB-IoT architectural to realize their full potential. In this paper, we construct a novel analytical model to estimate the random access performance of 5G NB-IoT network from an individual perspective. Then, based on the individual-based analytical model, we propose to decouple random access control management in 5G NB-IoT networks into coverage class adaptation and congestion control coordination, so that the critical issues of traffic burstiness and network scalability can be tackled. In particular, the random access control is progressively coordinated through a centralized SDN controller, and the coverage class adaptation can be achieved gradually through IoT devices under the guidance of the SDN controller. As a result, bursty traffics are equally dispersed among coverage classes, and the requirement of network control on complexity can be substantially alleviated, simplifying access control management and enhancing network scalability in a massive deployed NB-IoT networks. Case studies show that our SDN-enabled network control is effective for timely load balance.

With the emergence of cloud computing and several ultra-high bitrate data center applications, hybrid E/O switching intra-datacenter network (HS-IDCN) has become an integral architecture of current and future data centers. To meet the diverse and heterogeneous performance requirements of HS-IDCNs, people have considered traffic prediction as a promising solution to ensure effective and flexible traffic scheduling. However, the low accuracy of existing deep learning-based prediction approaches, which cannot fully extract the features of burst traffic, directly restricts the efficiency of traffic scheduling. In view of this, this study considers the spiking neural networks that can predict high burstiness and heterogeneous traffic to further improve the efficiency of traffic scheduling. We first propose a supervised spiking neural network (s-SNN) framework for high accuracy traffic prediction in HS-IDCNs. A traffic prediction-based traffic scheduling (TP-TS) algorithm for HS-IDCNs is then introduced by considering the prediction results of s-SNN. The s-SNN framework can enhance the extraction ability of burst traffic features in a supervised fashion by mimicking the multi-synaptic mechanism of biological neuron system. The efficiency and feasibility of s-SNN are verified on the brain model simulator. The performance of TP-TS is also evaluated in terms of resource utilization and path blocking probability, compared with other scheduling schemes.

Computation demands on mobile and edge devices are increasing dramatically. Mobile devices, such as smart phones, incorporate a large number of dedicated accelerators and fixed-function hardware blocks to deliver the required performance and power efficiency. Due to the heterogeneous nature of these devices, they feature vastly larger design spaces than traditional systems featuring only a CPU. Currently, academia struggles to fully evaluate such heterogeneous systems on chip due to the limited access and availability of proprietary workloads. To address these challenges, we propose Mocktails: a methodology to synthetically recreate the varying spatio-temporal memory access behaviour of proprietary heterogeneous compute devices. We focus on capturing the interspersed address streams of the workload and the burstiness of the injection process for proprietary compute devices commonly found in mobile systems. We evaluate Mocktails in simulation with proprietary memory traces of IP blocks. Mocktails accurately recreates the dynamic behaviour of memory access scheduling for memory controller metrics including read row hits (at most 7.3% error) and write row hits (at most 2.8% error). Architects can use Mocktails in their simulations as a substitute for a proprietary compute device, making the tool a useful conduit between industry and academia.

The E-model is the only recommended parametric tool for planning the quality of speech communication services, and its fullband version has recently been standardized by the International Telecommunication Union, ITU-T. In this paper, we analyze and extend the model by comparing its predictions for random and bursty packet loss as well as for delay to the results of a signal-based model, POLQA, as well as to the result of a subjective conversation test. The analysis shows that, by extending the fullband model to account for the burstiness, a reasonable prediction accuracy can be reached for random as well as bursty loss. The results are discussed and limitations of the current model are pointed out.

Spectrum prediction technology is able to make predictions about future use of the spectrum based on historical spectrum sensing data. ISM frequency bands for industrial, scientific and medical are part of the spectrum resources available to anyone without authorization, which makes this frequency band’s spectrum dynamic with highly burstiness, and difficult to discover the spectrum pattern and predict. In this paper, data analysis and feature mining are carried out on the real collected Wi-Fi data. According to the features of Wi-Fi data, a classification prediction framework based on LSTM model is developed. This algorithm framework can establish several classification predictors according to the features of Wi-Fi data mining and effective classification algorithms. In the process of prediction, the corresponding predictor is selected dynamically according to the characteristics of the history data. Experimental results with real data show that this prediction framework can effectively improve the accuracy of spectrum prediction.

Spectrum prediction has recently gained a lot of attention due to its extensive applications in cognitive radio networks. However, most of the related research assumed that the spectrum occupancy pattern is time-invariant, which limits the performance of proposed prediction methods for the high burst frequency bands, e.g. the ISM (Industrial, Scientific, Medical) bands. In order to improve the prediction accuracy for them, this paper first analyzes the characteristics of the collected real WiFi data in 2.4GHz ISM band, and shows the burstiness of the band from multiple aspects. Based on the analysis, we then propose a MultiLayer Perceptron based Reinforcement Learning (RLMLP) method which could adaptively select the corresponding predicting action according to the state to which any piece of data belongs. The state space of the method consists of multiple data categories that are determined by the results of feature partitioning, while the action space is composed of multiple MLPs with the same mapping structure. Finally, numerical results on the collected data show that the proposed RLMLP method is significantly better than the state-of-the-art algorithms in terms of the prediction performance.

In recent years, the amounts of network traffic targeted towards cloud data centers have fluctuated based on user requests. This traffic is bursty and requires a high degree of attention. Due to the variable nature of this traffic, some requests need to be re-allocated on-the-fly. Such circumstances result in performance degradations due to resource management. As appropriate solutions can be proposed only based on understanding the workload and the environment, Reinforcement Learning (RL) is a strategy that is predominantly used. Further, it has been shown that the Poisson arrival rates do not capture real-world burstiness. Thus, we mainly have a two-fold problem to address: (i) the traffic requires a new modelling approach that can characterize the burstiness, and (ii) balancing the load that can maximize the reward to the provider in such circumstances. In this paper, we propose a novel, yet simple traffic modelling that enables burst detection based on an index policy. We show that the throughput constraints play a crucial role in scheduling and our proposed RL technique produces reliable results in such a scenario. Our RL algorithm decides what instance of the request traffic needs to be processed so that the cloud provider can maximize its profit and the decisions made in hindsight are non-biased. We compare the proposed policy with two state-of-the-art approaches and draw key inferences as to why an index policy performs better in scenarios that demand RL. We observe over five times shorter average wait times while bursty workload crosses a saturation limit of 150% compared to conventional policies.

DDoS attacks plague the availability of online services today, yet like many cybersecurity problems are evolving and non-stationary. Normal and attack patterns shift as new protocols and applications are introduced, further compounded by burstiness and seasonal variation. Accordingly, it is difficult to apply machine learning-based techniques and defences in practice. Reinforcement learning (RL) may overcome this detection problem for DDoS attacks by managing and monitoring consequences ; an agent’s role is to learn to optimise performance criteria (which are always available) in an online manner. We advance the state-of-the-art in RL-based DDoS mitigation by introducing two agent classes designed to act on a per-flow basis, in a protocol-agnostic manner for any network topology. This is supported by an in-depth investigation of feature suitability and empirical evaluation. Our results show the existence of flow features with high predictive power for different traffic classes, when used as a basis for feedback-loop-like control. We show that the new RL agent models can offer a significant increase in goodput of legitimate TCP traffic for many choices of host density.

Network traffic is difficult to characterize due to its random, bursty nature. Even if a traffic source could be fit to a stochastic model with reasonable accuracy, analysis of end-to-end network performance metrics for such traffic models is generally intractable. In prior work, an approach to characterize traffic burstiness using a bound based on the class of phase-type distributions was proposed. Such phase-type bounds could be applied in conjunction with stochastic network calculus to derive probabilistic end-to-end delay bounds for a traffic stream. In this paper, we focus on the problem of estimating a tight phase-type burstiness bound for a given traffic trace. We investigate a method based on least squares and another based on the expectation-maximization algorithm. Our numerical results compare the two approaches in the scenario of a heavy-tailed M/G/1 queue. We find that while both methods are viable approaches for deriving phase-type bounds on traffic burstiness, the least squares approach performs better, particularly when a tail limit is imposed.1

This paper proposes a novel haptic packet rate control scheme for networked teleoperation systems which improves the state-of-the-art perceptual deadband-based haptic data reduction approach. The proposed method uses a peak-suppressing adaptive deadband to mitigate the burstiness of the haptic traffic and to control the packet rate. Experimental results show that our method is able to reduce the overall network load by reducing the burstiness of the haptic data transmission. It can also stabilize haptic packet rate beneath a certain transmission limit while preserving a high quality of remote interaction.

Network Functions Virtualization (NFV) has recently gained much popularity in the research scene for the flexibility and programmability that it will bring with the software implementation of network functions on commercial off-the-shelf (COTS) hardware. To substantiate its roll out, a number of issues (e.g., COTS’ inherent performance and energy efficiency, virtualization overhead, etc.) must be addressed, in a scalable and sustainable manner. Numerous works in the scientific literature manifest the strong correlation of network key performance indicators (KPIs) with the burstiness of the traffic. This paper proposes a novel model-based analytics approach for profiling virtualized network function (VNF) workloads, towards real-time estimation of network KPIs (specifically, power and latency), based on an MX/G/1/SET queueing model that captures both the workload burstiness and system setup times (caused by interrupt coalescing and power management actions). Experimental results show good estimation accuracies for both VNF workload profiling and network KPI estimation, with respect to the input traffic and actual measurements, respectively. This demonstrates that the proposed approach can be a powerful tool for scalable and sustainable network/service management and orchestration.=-1.

The information that short-lived TCP flows provide on bandwidth estimation may benefit adaptive video streaming applications or may contribute towards the success of new TCP versions. To estimate the available bandwidth active probing techniques may be used, but they cause an intrusive effect on the network affecting TCP performance. Therefore, passive measurements are favored, though capturing traffic traces comes with its own challenges. In this paper, we use the feedback provided by TCP acknowledgments and perform the estimation from sender-side measurements only. However, difficulties arise when in the presence of discrete random cross traffic, multiple tight links, packet losses, and inaccurate timestamping in general, the acknowledgment packet gaps get distorted. To deal with noise-afflicted packet gaps, we consider a machine-learning approach, specifically the neural network, for estimating the available bandwidth. We also apply the neural network under a variety of notoriously difficult conditions that have not been included in the training, such as multiple tight links, heavy cross traffic burstiness and packet losses. We compare the performance of our proposed method with a state-of-the-art model-based technique, where our neural network approach shows improved performance.

The emerging high-speed storage technologies and the ever-growing need for data storage are placing increasing pressure on data center networks. In this paper, we develop novel mechanisms for dynamically deciding when to move storage chunks or alter the number of active copies to alleviate congestion during high traffic episodes and to enable traffic consolidation (and hence network energy savings) during low traffic periods. Both actions are essential due to increasing burstiness of data center traffic. Using extensive simulations with modified NS3 network simulator we provide deep insights into the migration vs. replication tradeoff. We also show that under traffic bursts, a careful replication can provide up to 47% improvement in latency with read-dominated traffic, and strategic data migration to a low utilized node can provide up to 22% improvement in delays with write dominated traffic.

Communication links in low-power wireless visual sensor networks (WVSNs) are subject to short-term and long-term noise variations. These variations can cause a WVSN to exhibit prolonged or periodic transitional or bursty transmission performance. In this paper, we present our work on how to generate noise traces that simulate real-world transitional and bursty network behavior in TOSSIM. We develop a toolset called BurstyGen for TOSSIM which can facilitate WVSN protocol designers and application developers to better understand WVSN performance under these conditions. BurstyGen allows users to model both short time-scale and long time-scale variations in WVSN noise environments for the simulation and testing of WVSN system algorithms and protocols.

The paper addresses the problem of visual feature selection in content-based image retrieval (CBIR). We propose to study two strategies: the first one is using visual saliency, that selects the most salient features of the image and the second one exploits burstiness, that detects and processes the repeated visual elements in the image. To detect and describe the visual features in images, we rely on a deep local features approach based on convolutional neural network. The two strategies are evaluated for image retrieval on different datasets, according to two criteria: quality of retrieval and volume of manipulated features.

This paper addresses histogram burstiness, defined as the tendency of histograms to feature peaks out of proportion with their general distribution. After highlighting the impact of this growing issue on computer vision problems and the need to preserve the distribution information, we introduce a new normalization based on a Gaussian fit with a pre-defined variance for each datum that suppresses burst without adversely affecting the distribution. Experimental results on four public datasets show that our normalization scheme provides a staggering performance boost compared to other normalizations, even allowing Gaussian-normalized Bag-of-Words to perform similarly to intra-normalized Fisher vectors.

An accurate and fast estimation of the available bandwidth in a network with varying cross-traffic is a challenging task. The accepted probing tools, based on the fluid-flow model of a bottleneck link with first-in, first-out multiplexing, estimate the available bandwidth by measuring packet dispersions. The estimation becomes more difficult if packet dispersions deviate from the assumptions of the fluid-flow model in the presence of non-fluid bursty cross-traffic, multiple bottleneck links, and inaccurate time-stamping. This motivates us to explore the use of machine learning tools for available bandwidth estimation. Hence, we consider reinforcement learning and implement the single-state multi-armed bandit technique, which follows the e-greedy algorithm to find the available bandwidth. Our measurements and tests reveal that our proposed method identifies the available bandwidth with high precision. Furthermore, our method converges to the available bandwidth under a variety of notoriously difficult conditions, such as heavy traffic burstiness, different cross-traffic intensities, multiple bottleneck links, and in networks where the tight link and the bottleneck link are not same. Compared to the piece-wise linear network a model based direct probing technique that employs a Kalman filter, our method shows more accurate estimates and faster convergence in certain network scenarios and does not require measurement noise statistics.

Transmission Control Protocol with Network Coding (TCP/NC) has been a potential proposal to improve the transmission performance of TCP in lossy networks while it still takes advantage of congestion and retransmission control of TCP for reliable end-to-end data transfer. TCP/NC uses additional sub-layer called Network Coding layer below the TCP layer to handle packet losses without sensed by the TCP layer. In our previous work, we introduced some variants of TCP/NC such as TCP/NCwLRLBE (TCP/NC with Loss Rate and Loss Burstiness Estimation) which can improve the retransmission and adapt to the changing of the channel. However, it is just evaluated on the simple packet loss conditions, e.g., random loss or Gilbert loss. On the other hand, reliable end-to-end data transfer in diverse IoT (Internet of Things) environments (e.g., Power Line Communication - PLC) in a cost-efficient manner is an emerging challenge. Therefore, in this paper, we investigate the characteristics of the TCP/NC on PLC networks, to have a more in-depth evaluation of its performance and applicability in a practical PLC environment with many conditions causing packet loss such as noise variants, distance, etc. The result of our simulation on ns-3 (Network Simulation 3) shows that TCP/NC can work well on PLC environment compared to the traditional protocols such as TCP NewReno with Selective Acknowledgment (SACK) and TCP Westwood+.

Digital single-flux quantum (SFQ) technology promises to meet the demands of ultra low power and high speed computing needed for future exascale supercomputing systems. The combination of ultra high clock frequencies, gate-level pipelines, and numerous sources of variability in SFQ circuits, however, make low-skew global clock distribution a challenge. This motivates the support of multiple independent clock domains and related clock domain crossing circuits that enable reliable communication across domains. Existing J-SIM simulation models indicate that setup violations can cause clock-to-Q increases of up to 100%. This paper first shows that naive SFQ clock domain crossing (CDC) first-in-first-out buffers (FIFOs) are vulnerable to these delay increases, motivating the need for more robust CDC FIFOs. Inspired by CMOS multi-flip-flop asynchronous FIFO synchronizers, we then propose a novel 1-bit metastability-resilient SFQ CDC FIFO that simulations show delivers over a 1000 reduction in logical error rate at 30 GHz. Moreover, for a 10-stage FIFO, the Josephson junction (JJ) area of our proposed design is only 7.5% larger than the non-resilient counterpart. Finally, we propose design guidelines that define the minimal FIFO depth subject to both throughput and burstiness constraints.

Previous studies have observed that TCP pacing evenly spacing out packets-minimizes traffic burstiness, reduces packet losses, and increases throughput. However, the main drawback of pacing is that the number of flows and the bottleneck link capacity must be known in advance. With this information, pacing is achieved by manually tuning sender nodes to send at rates that aggregate to the bottleneck capacity. This paper proposes a scheme based on programmable switches by which rates are dynamically adjusted. These switches store the network’s state in the data plane and notify sender nodes to update their pacing rates when the network’s state changes, e.g., a new flow joins or leaves the network. The scheme uses a custom protocol that is encapsulated inside the IP Options header field and thus is compatible with legacy switches (i.e., the scheme does not require all switches to be programmable). Furthermore, the processing overhead at programmable switches is minimal, as custom packets are only generated when a flow joins or leaves the network. Simulation results conducted in Mininet demonstrate that the proposed scheme is capable of dynamically notifying hosts to adapt the pacing rate with a minimum delay, increasing throughput, mitigating the TCP sawtooth behavior, and achieving better fairness among concurrent flows. The proposed scheme and preliminary results are particularly attractive to applications such as Science DMZ, where typically a small number of large flows must share the bandwidth capacity.

The bursty nature of network traffic makes it difficult to characterize accurately, and may give rise to heavy-tailed queue distributions within the network. Building on prior work in stochastic network calculus, we propose traffic burstiness bounds based on the class of phase-type distributions and develop an approach to estimate the parameter of such bounds using the expectation-maximization (EM) algorithm. By limiting the tail of the burstiness bound, our approach achieves a better fit of the phase-type distribution to the empirical data from heavy-tailed traffic. The proposed tail-limited phase-type burstiness bounds fall within the framework for stochastic network calculus based on generalized stochastically bounded burstiness. We demonstrate the effectiveness of the proposed methodology with a numerical example involving a heavy-tailed M/G/1 queue.1

Transmission Control Protocol with Network Coding (TCP/NC) is studied to improve the goodput performance of the standard TCP in lossy networks (e.g., wireless networks). TCP/NC uses additional sub-layer called Network Coding layer below TCP layer to handle packet losses without sensed by TCP layer. Basically, n+k combination packets combined from n original packets are sent by the source. When the sink receives enough n combination packets, the sink can calculate n original packet even though k ' packets are lost where k ' is less than or equal to k. Most versions of TCP/NC consider the loss in only the direction of sending data; however, the loss in the reverse direction, i.e., the direction of sending an acknowledgment, affects seriously to goodput performance, especially in TCP/NC with an automatic estimation of Network Coding parameters n and k) case. In this paper, we propose a new scheme for the bi-directional loss issue. The result of our simulation on ns-3 (Network Simulation 3) shows that the proposed scheme can work well when loss happens in both directions compared to the TCP NewReno and our previously proposed, TCP/NC with Loss Rate and Loss Burstiness Estimation (TCP/NCwLRLBE)., 18th International Conference on Electronics, Information, and Communication (ICEIC 2019), January 22 - 25, 2019, Pullman Auckland Hotel, New Zealand

In the Internet of Things (IoT), wireless sensor networks are often paired with machine learning frameworks to deliver applications of high societal impact and support critical infrastructures. In this context, this paper investigates the relationship between network reliability and the reliability of the machine learning framework in terms of prediction accuracy. Our experimental analysis leverages six data sets of various degrees of information redundancy and considers four machine learning algorithms that are commonly used for classification. In turn, packet loss is inserted in the raw input data, emulating various networking loss patterns in terms of burstiness. The experimental results consistently demonstrate a non-linear relationship between the reliability of the network and the accuracy of the machine learning classifier, indicating that not all data packets are equally valuable to the application performance. We conclude with recommendations for IoT practitioners and IoT system designers.

Because of the burstiness and uncertainty of network, the prediction for short-term network traffic is a difficult problem. This paper proposes a real-time network traffic prediction model based on Long Short-Term Memory (LSTM) neural network. The loss function of LSTM network is modified to enhance the robustness of the prediction model. Different from the traditional LSTM model, the proposed model is continually updated with the arrival of new traffic. The experimental results show that the proposed model performs better on prediction accuracy than other models constructed with Support Vector Regression and Back Propagation neural network.

Online social streams has become a prominent channel for facilitating new sources of information and quick communication links, specifically during mass emergencies. The task of event burst detection focuses on monitoring Twitter streams, to identify emergent patterns based on tweet arrival rate. The vital features of rising sharply in frequency for modelling an event in text streams leads to burst activity. In this paper, a statistical approach is framed for Twitter streaming channel with a sudden change in the frequency of tweet showing a log-normal distribution of data arrival $\gamma$. The burstiness of event is framed as the Z-score of the tweet arrival rate $\lambda$ which corresponds to the frequency of tweets delimited in the time window w. Moreover, the proposed approach identify events which generates a stratified representation of the set of keyword bursts. The experimental outcome determines a hierarchical composition of event detection on the diversified stream by choosing optimal parameters.