• Cross-Layer Cooperative Protocols for Wireless Networks
• Blind MIMO System Identification
• High-Speed Network Traffic Modeling
• Channel Estimation for Communications
• Higher-Order Spectra Analysis
• Biomedical Signal Processing
• Carrier Frequency Offsets Estimation
• Selected Senior Design Projects

High Speed Network Traffic Modeling

Extensive data studies indicate that traffic in high-speed communication networks has long-memory and heavy-tailed (impulsive) characteristics. With the rising popularity of multimedia applications over networks, these properties of the traffic are only likely to become more dominant, posing unique new challenges to designers of network systems and protocols. Traditional teletraffic theory cannot capture these traffic characteristics. During the last few years, significant research results have been proposed on models that capture self-similarity of traffic. These models, however, are inadequate for predicting queuing performance, delays and buffer dimensioning since the implications of the combination of self-similarity and impulsiveness on queueing performance can be dramatically different from that of self-similarity alone. We are developing models by taking into account real traffic dynamics that capture the data impulsiveness as well as self-similarity.

 

Extensive results on self-similar traffic modeling have emerged in literature. See, for example, [Leland 95]. Various analytical models have been proposed to model the high-speed communication network traffic. However, these models were constructed upon the Gaussian marginal distribution assumption, which departs far away from actual measured traffic data. The next graph illustrates the histogram of the actual traffic data (1st row) and synthesized trace from the well-established On/Off model (2nd row) [Willinger 97].



From this graph it becomes evident that the real traffic is clearly non-Gaussian, a characteristic that is not captured by the On/Off model.

The non-Gaussian (heavy-tailed) marginal distribution of traffic can challenge traditional network infrastructure design strategies, such as buffer dimensioning, packet routing and scheduling etc. Thus, an accurate model for the high-speed communication networks should be geared towards capturing heavy-tailed marginal characteristics as well the self-similarity. We have recently proposed and analyzed a new model, the Extended Alternating Fractal Renewal Process (EAFRP) to model network traffic. Synthesized traffic based of the EAFRP (see figure below) appears to follow well actual traffic data.

Features