An official website of the United States government
Here’s how you know
Official websites use .gov
A .gov website belongs to an official government organization in the United States.
Secure .gov websites use HTTPS
A lock (
) or https:// means you’ve safely connected to the .gov website. Share sensitive information only on official, secure websites.
Understanding the Correlation Structure of Network Traffic
Published
Author(s)
Kevin L. Mills, J Yuan
Abstract
This paper aims to improve current understanding of the correlation structure of traffic carried over large networks, such as the Internet. To achieve this aim we use simulation, adopting a methodology of homogenization to achieve a sufficiently large model with well-understood parameters. Such a model enables systematic study of fundamental causalities arising from interactions among factors in the application layer, the transport layer, and the network structure. Focusing the model on describing comparative rather than absolute behavior, we undertake a systematic search, using wavelet-based analysis, to identify and understand significant phenomena influencing the correlation structure of network traffic. We find significant interaction between offered traffic and network congestion, andwe conclude that the correlation structure of network traffic should be controllable by modulating available resources. We also find thatcongestion-control mechanisms affect the characteristics of timescale dynamics in network traffic. We illustrate how variability in networkstructure leads to changes in correlation structure. In particular, we find a similar correlation structure to that seen for measured Internet traffic may arise in very large networks, even without high user variability. Our findings imply that observed traffic characteristics might be a combined effect of many factors, including user behavior, transport mechanism, and network structure. Our results suggest that while searching for invariants from empirical observations, one must take care to identify combined effects. Finally, results discussed in this paper suggest that network scale has a more general influence on correlation structure than other properties, such as heavy-tailed file sizes.
Mills, K.
and Yuan, J.
(2002),
Understanding the Correlation Structure of Network Traffic, IEEE-Acm Transactions on Networking, [online], https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=151109
(Accessed November 1, 2024)