Predicting the routing paths between any given pair of Autonomous Systems (ASes) is very useful in network diagnosis, traffic engineering, and protocol analysis. Existing methods address this problem by resolving the best path with a snapshot of BGP (Border Gateway Protocol) routing tables. However, due to route deficiencies, routing policy changes, and other causes, the best path changes over time. Consequently, existing methods for path prediction fail to capture route dynamics. To predict AS-level paths in dynamic scenarios (e.g. network failures), we propose a per-neighbor path ranking model based on how long the paths have been used, and apply this routing model to extract each AS's route choice configurations for the paths observed in BGP data. With route choice configurations to multiple paths, we are able to predict the path in case of multiple network scenarios. We further build the model with strict policies to ensure our model's routing convergence; formally prove that it converges; and discuss the path prediction capturing routing dynamics by disabling links. By evaluating the consistency between our model's routing and the actually observed paths, we show that our model outperforms the state-of-the-art work [4].
Shen SU
Harbin Institute of Technology
Binxing FANG
Harbin Institute of Technology
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copy
Shen SU, Binxing FANG, "Towards Route Dynamics in AS-Level Path Prediction" in IEICE TRANSACTIONS on Communications,
vol. E99-B, no. 2, pp. 412-421, February 2016, doi: 10.1587/transcom.2015EBP3203.
Abstract: Predicting the routing paths between any given pair of Autonomous Systems (ASes) is very useful in network diagnosis, traffic engineering, and protocol analysis. Existing methods address this problem by resolving the best path with a snapshot of BGP (Border Gateway Protocol) routing tables. However, due to route deficiencies, routing policy changes, and other causes, the best path changes over time. Consequently, existing methods for path prediction fail to capture route dynamics. To predict AS-level paths in dynamic scenarios (e.g. network failures), we propose a per-neighbor path ranking model based on how long the paths have been used, and apply this routing model to extract each AS's route choice configurations for the paths observed in BGP data. With route choice configurations to multiple paths, we are able to predict the path in case of multiple network scenarios. We further build the model with strict policies to ensure our model's routing convergence; formally prove that it converges; and discuss the path prediction capturing routing dynamics by disabling links. By evaluating the consistency between our model's routing and the actually observed paths, we show that our model outperforms the state-of-the-art work [4].
URL: https://global.ieice.org/en_transactions/communications/10.1587/transcom.2015EBP3203/_p
Copy
@ARTICLE{e99-b_2_412,
author={Shen SU, Binxing FANG, },
journal={IEICE TRANSACTIONS on Communications},
title={Towards Route Dynamics in AS-Level Path Prediction},
year={2016},
volume={E99-B},
number={2},
pages={412-421},
abstract={Predicting the routing paths between any given pair of Autonomous Systems (ASes) is very useful in network diagnosis, traffic engineering, and protocol analysis. Existing methods address this problem by resolving the best path with a snapshot of BGP (Border Gateway Protocol) routing tables. However, due to route deficiencies, routing policy changes, and other causes, the best path changes over time. Consequently, existing methods for path prediction fail to capture route dynamics. To predict AS-level paths in dynamic scenarios (e.g. network failures), we propose a per-neighbor path ranking model based on how long the paths have been used, and apply this routing model to extract each AS's route choice configurations for the paths observed in BGP data. With route choice configurations to multiple paths, we are able to predict the path in case of multiple network scenarios. We further build the model with strict policies to ensure our model's routing convergence; formally prove that it converges; and discuss the path prediction capturing routing dynamics by disabling links. By evaluating the consistency between our model's routing and the actually observed paths, we show that our model outperforms the state-of-the-art work [4].},
keywords={},
doi={10.1587/transcom.2015EBP3203},
ISSN={1745-1345},
month={February},}
Copy
TY - JOUR
TI - Towards Route Dynamics in AS-Level Path Prediction
T2 - IEICE TRANSACTIONS on Communications
SP - 412
EP - 421
AU - Shen SU
AU - Binxing FANG
PY - 2016
DO - 10.1587/transcom.2015EBP3203
JO - IEICE TRANSACTIONS on Communications
SN - 1745-1345
VL - E99-B
IS - 2
JA - IEICE TRANSACTIONS on Communications
Y1 - February 2016
AB - Predicting the routing paths between any given pair of Autonomous Systems (ASes) is very useful in network diagnosis, traffic engineering, and protocol analysis. Existing methods address this problem by resolving the best path with a snapshot of BGP (Border Gateway Protocol) routing tables. However, due to route deficiencies, routing policy changes, and other causes, the best path changes over time. Consequently, existing methods for path prediction fail to capture route dynamics. To predict AS-level paths in dynamic scenarios (e.g. network failures), we propose a per-neighbor path ranking model based on how long the paths have been used, and apply this routing model to extract each AS's route choice configurations for the paths observed in BGP data. With route choice configurations to multiple paths, we are able to predict the path in case of multiple network scenarios. We further build the model with strict policies to ensure our model's routing convergence; formally prove that it converges; and discuss the path prediction capturing routing dynamics by disabling links. By evaluating the consistency between our model's routing and the actually observed paths, we show that our model outperforms the state-of-the-art work [4].
ER -