Characterizing and Modeling WAN Egress Traffic
No Thumbnail Available
Date
2018-05
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
AAU
Abstract
Computer networks exhibit complex characteristics due to the heterogenous nature of traffic
running through the network. This makes the design of reliable networks and network services
difficult. To have a design of robust and reliable networks, a detailed understanding of traffic
characteristics of the network is needed which will lead to distinguish the traffic model it fits.
In this paper, it is showed that the WAN egress traffic possess self-similar characteristics, using
different mathematical techniques. And also, the presence of long memory in WAN egress traffic
is shown by the Autocorrelation Function of the trace. Additionally, it is showed that one of the
self-similar long memory models, Fractional Auto-Regressive Integrated Moving Average
(FARIMA) model, best capture the collected WAN traffic data. To model the traffic data first
stationarity was tested using Augmented Dickey Fuller (ADF) test. The AR and MA terms of the
model are estimated using the ACF and PACF plot. To test the model, Autocorrelation function is
used, and it is found that the Autocorrelation function of the approximated data has a resemblance
to the Autocorrelation function of the collected data.
Description
Keywords
WAN, Self-similarity, Long Range Dependence, Traffic models, FARIMA