Cui Yidong

Network traffic prediction by a wavelet-based combined model

Network traffic prediction models can be grouped into two types, single models and combined ones. Combined models integrate several single models and thus can improve prediction accuracy. Based on wavelet transform, grey theory, and chaos theory, this paper proposes a novel combined model, wavelet–grey–chaos (WGC), for network traffic prediction. In the WGC model, we develop a time series decomposition method without the boundary problem by modifying the standard a trous algorithm, decompose the network traffic into two parts, the residual part and the burst part to alleviate the accumulated error problem, and employ the grey model GM(1,1) and chaos model to predict the residual part and the burst part respectively. Simulation results on real network traffic show that the WGC model does improve prediction accuracy.

UGM based internet traffic prediction

Internet traffic exhibits complex temporal multi-scale properties. The properties greatly affect the predicting method based upon grey theory. The general GM(1,1) model will lose effectiveness if the modeling length approaches to the traffic variation period. We proposed a new method based upon undulating grey model (UGM) to predict the internet traffic. Simulated with traffic data from Abilene network, the new method is verified to perform better than the general GM(1,1) model.

The reasearch of hierarchy model for ad hoc network monitoring based on clustering

Ad hoc network has dynamic topology and large-scale nodes, for the purpose of monitoring and measurement in Ad hoc network, this paper proposes a new frame method of hierarchy model for monitoring, i.e., clusterhead selection mechanism. The mechanism designed a weight formula to handle the comprehensive update considering several factors such as distance, mobile rate, energy during the model establishment period, and three kinds of partial update means in real-time attributes management. It enhances load balance and structure stability in monitor hierarchy model and reduces calculation cost for unnecessary update. The simulation results show that the model using clusterhead selection mechanism was stable and efficient, which could apply to muti-scale networks, achieve conveniently and flexible deployment in Ad hoc network measurement system.

Chaotic Control of Network Traffic

A method of chaotic control on network traffic is presented. By this method, the chaotic network traffic can be controlled to a pre-assigned equilibrium point according to chaotic prediction and the largest Lyapunov exponent of the traffic on congested link is reduced, thereby the probability of traffic burst and network congestion can be reduced. Numerical examples show that this method is effective.

An approach to evaluate network performance based on hop-by-hop latencies

Understanding network performance especially in link level is becoming increasingly important for network management. This paper puts forward an approach to evaluate network performance based on hop-by-hop latencies from three different levels of granularity including interface granularity, city granularity and AS granularity. We use a set of large scaled traceroute data, from a number of sources which are PlanetLab nodes to other random selected destinations to accomplish the experimental study. The analysis results show that this approach can carry out network performance evaluation in a flexible and effective way.

Research on improved INSIGNIA based on network measurement

Due to the difficulty of QOS assurance for MANET(Mobile Ad hoc Network) in which the network topology changes dynamically, we improve the INSIGNIA protocol and proposes a method based on network measurement to provide better QOS assurance. This paper introduces the design principle and implementation of this method in details, and then introduces the system model of this method which is designed and implemented in OPNET simulation environment, describes the experiment process of verifying the feasibility and effectivily of this system model through simulation result.

A novel approach to link utilization measurement

In this paper a novel approach to link utilization measurement is proposed. In the network, the source workstation sends IP packets (namely, probes) that have the time stamp options being set to an arbitrary ICMP-enabled destination. Each router that the probes traverse across will record a timestamp in the packets. The timestamps are collected and further processed to gain the link utilization. By analyzing the model of the single hop delay, it is observed that the packet processing delay, propagation delay, the clock skew greatly affect the timestamps. So, the LUS (Link Utilization Statistic) algorithm is proposed to remove the disturbance. The LUS algorithm computes the queuing delay of each probe by processing multiple timestamps jointly. Then being quantized and smoothed with slide window, the queuing delay is transformed into link utilization. This approach is verified through simulation.