Time-Driven vs packet-driven: a deep study on traffic sampling

Authors:
Xiaoxin Shao;Tao He;Shijin Kong;Changqing An;Xing Li
Affiliations:
Department of Electronic Engineering, Tsinghua University, Beijing, P.R China;China Education and Research Network, Beijing, P.R China;Department of Electronic Engineering, Tsinghua University, Beijing, P.R China;China Education and Research Network, Beijing, P.R China;Department of Electronic Engineering, Tsinghua University, Beijing, P.R China
Venue:
ICOIN'06 Proceedings of the 2006 international conference on Information Networking: advances in Data Communications and Wireless Networks
Year:
2006

Citing 6
Cited 0

Internet traffic characterization

Internet traffic characterization
New directions in traffic measurement and accounting

IMW '01 Proceedings of the 1st ACM SIGCOMM Workshop on Internet Measurement
Charging from sampled network usage

IMW '01 Proceedings of the 1st ACM SIGCOMM Workshop on Internet Measurement
Design and Deployment of a Passive Monitoring Infrastructure

IWDC '01 Proceedings of the Thyrrhenian International Workshop on Digital Communications: Evolutionary Trends of the Internet
Building a better NetFlow

Proceedings of the 2004 conference on Applications, technologies, architectures, and protocols for computer communications
An In-Depth, Analytical Study of Sampling Techniques for Self-Similar Internet Traffic

ICDCS '05 Proceedings of the 25th IEEE International Conference on Distributed Computing Systems

Quantified Score

Hi-index	0.00

Visualization

Abstract

Traffic sampling technology has been widely deployed in front of many high-speed network applications to alleviate the great pressure on packet capturing Packet-driven sampling mechanism is believed better than time-driven one and no in-depth comparison is given to these two kinds of mechanisms In this paper, a systematic comparison is conducted on three sampling methods, 1/N packet-driven, 1/T time-driven and t/T time-driven samplings, with a result showing that t/T sampling achieves similar accuracy as 1/N sampling in most aspects, and surpasses 1/N sampling in estimating the interval time distribution Then we try to optimize performance of t/T sampling by tuning its parameters, and verify putative conclusions with both real and simulation traffic The experiment in a real measurement application also indicates that these two kinds of sampling mechanisms achieve similar online estimation performance.