Research Article
Analysis and Detection of Bottlenecks via TCP Footprints in live 3G Networks
@INPROCEEDINGS{10.4108/ICST.WIOPT2008.3256, author={Philipp Svoboda and Fabio Ricciato}, title={Analysis and Detection of Bottlenecks via TCP Footprints in live 3G Networks}, proceedings={4th International ICST Workshop On Wireless Network Measurement}, publisher={IEEE}, proceedings_a={WINMEE}, year={2008}, month={8}, keywords={Bandwidth Base stations Bit error rate Counting circuits Data mining Protocols Scattering Statistical analysis Telecommunication congestion control Telecommunication traffic}, doi={10.4108/ICST.WIOPT2008.3256} }- Philipp Svoboda
Fabio Ricciato
Year: 2008
Analysis and Detection of Bottlenecks via TCP Footprints in live 3G Networks
WINMEE
IEEE
DOI: 10.4108/ICST.WIOPT2008.3256
Abstract
In this paper we evaluate four different metrics for non intrusive bottleneck detection based on TCP counters. This work is based on the full TCP statistics recorded on five days spread over the last one and a half year within the core network of a mobile network operator in Austria. Scatterplots, so called “footprints”, were generated counting the number of packets and the number of retransmission for each user during the peak hours. Two of the datasets had a known capacity bottleneck in place. Based on those datasets we benchmarked the different metrics for the detection of a bottleneck event. We preprocessed the traces in order to remove the traffic increase. After this step all metrics were able to detect the special bottleneck case. Even traces separated for more than one year deliver a clear result. The performance of a PSNR metric was similar to the other metrics based on more sophisticated functions.
