Flávio Henrique Teles Vieira

A network traffic prediction approach based on multifractal modeling

This work extends the notion of the widely mentioned and used fractional Brownian traffic model in the literature. Extensive experimental investigations indicate that the proposed traffic model, named extended fractional Brownian traffic, can capture not only the self-similar properties, but also the inherent multifractal characteristics of those traffic flows found in modern communication networks. Additionally, the structure of this traffic model is taken into account in a traffic prediction algorithm that benefits from the more accurate traffic modeling. The experimental results clearly point out the advantages of using the proposed model in traffic modeling as well as in traffic prediction.

A novel network traffic predictor based on multifractal traffic characteristic

In this work, we propose a network traffic predictor based on a novel multifractal network traffic model. This multifractal traffic model extends the notion of the classical fractional Brownian traffic model proposed by Norros (Queueing Systems, vol.16, p.387-396, 1994), replacing the fractional Brownian motion (fBm) process by the multifractional Brownian motion (mBm) process. The network traffic is assumed to be modeled by the extended fractional Brownian traffic model, which is characterized by its Holder exponents. The value of the Holder exponent at a given time indicates the degree of the traffic burstiness at that time. Based on the mBm covariance structure, a mean-square error discrete-time k-step predictor is implemented. Tests carried out using real wide area network traffic traces proved the proposed traffic model and predictor effectiveness.

Optimal positioning of GPRS concentrators for minimizing node hops in smart grids considering routing in mesh networks

Communication systems based on RF Mesh are widely used in various applications of smart metering, an essential component of Smart Grids. The RF Mesh technology has a long history and proved to be highly reliable in remote telemetry and distributed control applications. This paper presents a methodology for optimization of ZigBee mesh networks as the communication technology support for a Smart Grid. More specifically, we propose an algorithm to determine the optimal location of GPRS devices concentrators applied to AMI (Advanced Metering Infrastructure), considering the traffic behavior in a Smart Grid. We validate the solution of GPRS concentrator positioning given by our algorithm by simulating the Smart Grid scenarios and collecting traffic statistics such as average delay and traffic load in the concentrator.

Loss probability estimation and control for OFDM/TDMA wireless systems considering multifractal traffic characteristics

In this paper, we analyze the queueing performance in terms of loss rate of an OFDM (orthogonal frequency-division multiplexing)/TDMA (time division multiplexing access) based wireless system taking into account the multifractal behavior of the wireless traffic flows. To this end, first, we show evidences of multifractal characteristics on wireless traffic traces. These findings motivated us to propose a traffic policing and control scheme based on a multifractal envelope process in order to maintain the traffic flows well-behaved, i.e., in accordance to the desired QoS parameters. Furthermore, by assuming a multifractal traffic model, we derive a data loss probability equation for wireless traffic flows that was applied to the OFDM/TDMA based wireless system. Simulations and comparisons to other methods were carried out in order to verify the efficiency of the proposed traffic policing scheme as well as of the loss probability estimation approach.

Estimation of backlog and delay in OFDM/TDMA systems with traffic policing using Network Calculus

Achieving quality of service (QoS) in OFDM based networks depends, among other factors, on mechanisms for traffic policing or control. Aggregate traffic in high speed networks can present multifractal characteristics such as bursts over a wide range of time scales, heavy tailed inter-arrival time densities, self-similarity and long-range dependence. These properties are followed by a degradation of quality of service for the traffic flows. In this paper, we firstly propose a policing algorithm based on multifractal traffic modeling, evaluating its performance in comparison to other algorithms. We present a more precise in general envelope modeling process in order to efficiently police the input traffic to the system. Further, we present an approach based on Network Calculus for estimating quality of service parameters, such as mean queue length (backlog) and delay. We analyze these two parameters for an OFDM/TDMA based simplified WiMAX system with and without the application of traffic modeling based policing algorithms. In other words, we propose an approach for evaluating some QoS parameters of an OFDM/TDMA system where the traffic is policed. We also evaluate the performance of the proposed algorithm in policing the traffic in an OFDM/TDMA system.

Estimation of Backlog and Delay in OFDM/ TDMA Systems with Traffic Policing Through Network Calculus

In this paper, we firstly propose a policing algorithm based on multifractal traffic modeling, evaluating its performance in comparison to other algorithms. Further, we present an approach based on Network Calculus for estimating quality of service parameters, such as mean queue length (backlog) and delay. We analyze these two parameters for an OFDM/TDMA based simplified Wimax system with and without the application of traffic modeling based policing algorithms. We also evaluate the performance of the proposed algorithm in policing the traffic in an OFDM / TDMA system.

Neural Network Based Approaches for Network Traffic Prediction

In this chapter, we review some learning strategies for neural networks such as MLP (Multilayer Perceptron), RBF (Radial Basis Function) and Recurrent Networks applied to computer network traffic prediction. That is, the considered neural networks and training algorithms are used to predict the traffic volume of a computer network. Some methods of improving the prediction performance of neural networks are also considered such as application of Wavelet Transform. We discuss about using the Wavelet Transform in supervised training of neural networks by decomposing the traffic process into approximation and detail processes. We present some results involving the application of the Orthogonal Least Squares (OLS) algorithm in RBF networks for traffic prediction. Regarding the Recurrent neural networks, we verify their traffic prediction performance when trained with the Extended Kalman Filter (EKF) and the RTRL (Real Time Recurrent Learning). Real network traffic traces are used in the simulations in order to verify the prediction performance of the neural network algorithms.

An adaptive fuzzy model using orthonormal basis functions based on multifractal characteristics applied to network traffic control

In this paper, we present an adaptive predictive orthonormal basis functions (OBF)-fuzzy model that considers the multifractal behavior of network traffic flows. To this end, we model the traffic flows using orthonormal basis functions obtained through multifractal analysis. We insert the orthonormal basis functions into a fuzzy model trained with an adaptive clustering algorithm. Further, we propose a predictive flow control scheme for broadband networks. Also, using the fuzzy model parameters, we derive an expression for the optimal traffic source rate. Comparisons to other predictive control schemes prove the efficiency of the proposed adaptive OBF-fuzzy based control and training algorithm.

Modeling of MPEG-4 Video Traffic Using a Multifractal Cascade with Autoregressive Multipliers

In this paper we propose a video traffic model based on a multifractal multiplicative cascade. In our proposal, we assume that the cascade multipliers can be efficiently described by autoregressive models in each cascade scale. This multiplicative cascade model with autoregressive process based multipliers is used for GOP (Group of Pictures) level traffic modeling as well as for frame level traffic modeling. Aiming to validate the model, we performed statistical and queue behavior analysis of the video traffic processes generated using the proposed model. We also verified the model performance in capturing the real MPEG-4 video traffic characteristics compared to other multifractal models.

A management system for PLC networks using SNMP Protocol

Access technologies for data transmission, such as xDSL, Wi-Fi and cable modem are widely used once they support high data transmission rates at low cost. Among these technologies, Power Line Communications (PLC) is an alternative promising solution. In order to get control over a telecommunication network operation, including PLC, it is necessary to use management techniques. Besides network control, these management techniques are intended to provide maximum knowledge of the system behavior and of involved devices. In this paper, we present a management solution model applied to PLC networks. In order to develop the proposed management system for PLC networks, we made use of a free software called Nagios. Practical evaluation tests showed that the proposed management model is scalable, reliable and can provide effective control of PLC networks in order to guarantee the required services to the users.