Chair for Network Architectures and Services Technical University of Munich (TUM) Correlating TTL data to network characteristics Final Talk BSc Informatics Till Wickenheiser Advisors: Quirin Scheitle, Oliver Gasser, Paul Emmerich, Felix von Eye April 27, 2016 Chair for Network Architectures and Services Department of Informatics Technical University of Munich (TUM) Till Wickenheiser – Correlating TTL data to network characteristics 1
Chair for Network Architectures and Services Technical University of Munich (TUM) Motivation Research questions Approach Implementation Discussion of the results Future Work Till Wickenheiser – Correlating TTL data to network characteristics 2
Chair for Network Architectures and Services Technical University of Munich (TUM) Motivation ◮ Time to Live(TTL) behavior strongly influenced by network structure Till Wickenheiser – Correlating TTL data to network characteristics 3
Chair for Network Architectures and Services Technical University of Munich (TUM) Motivation ◮ Time to Live(TTL) behavior strongly influenced by network structure ◮ Can knowledge about the network characteristics be used to predict incoming TTL values? Till Wickenheiser – Correlating TTL data to network characteristics 3
Chair for Network Architectures and Services Technical University of Munich (TUM) Research questions ◮ Is there a correlation between BGP data and hop count of incoming packets? Till Wickenheiser – Correlating TTL data to network characteristics 4
Chair for Network Architectures and Services Technical University of Munich (TUM) Research questions ◮ Is there a correlation between BGP data and hop count of incoming packets? ◮ Is there a correlation between RTT of observed hosts and hop counts? Till Wickenheiser – Correlating TTL data to network characteristics 4
Chair for Network Architectures and Services Technical University of Munich (TUM) Research questions ◮ Is there a correlation between BGP data and hop count of incoming packets? ◮ Is there a correlation between RTT of observed hosts and hop counts? ◮ Is there a correlation between Geo location data and incoming hop count? Till Wickenheiser – Correlating TTL data to network characteristics 4
Chair for Network Architectures and Services Technical University of Munich (TUM) Approach Correlation of hop count and BGP data: ◮ Capture a sample of the Internet traffic using the already implemented TTL capturing framework (by Christian Sturm) Till Wickenheiser – Correlating TTL data to network characteristics 5
Chair for Network Architectures and Services Technical University of Munich (TUM) Approach Correlation of hop count and BGP data: ◮ Capture a sample of the Internet traffic using the already implemented TTL capturing framework (by Christian Sturm) ◮ Use the local BGP data as a model for the routing used for the traffic sample Till Wickenheiser – Correlating TTL data to network characteristics 5
Chair for Network Architectures and Services Technical University of Munich (TUM) Approach Correlation of hop count and BGP data: ◮ Capture a sample of the Internet traffic using the already implemented TTL capturing framework (by Christian Sturm) ◮ Use the local BGP data as a model for the routing used for the traffic sample ◮ Compare the hop count of packets to the length of AS paths using the corresponding BGP data Till Wickenheiser – Correlating TTL data to network characteristics 5
Chair for Network Architectures and Services Technical University of Munich (TUM) Implementation ◮ Create a longest prefix match lookup tree of entries of the BGP data (libbgpdump, pytricia) Till Wickenheiser – Correlating TTL data to network characteristics 6
Chair for Network Architectures and Services Technical University of Munich (TUM) Implementation ◮ Create a longest prefix match lookup tree of entries of the BGP data (libbgpdump, pytricia) ◮ Convert the TTL values to hop counts Till Wickenheiser – Correlating TTL data to network characteristics 6
Chair for Network Architectures and Services Technical University of Munich (TUM) Implementation ◮ Create a longest prefix match lookup tree of entries of the BGP data (libbgpdump, pytricia) ◮ Convert the TTL values to hop counts ◮ Assign an AS path to each entry in the capture file Till Wickenheiser – Correlating TTL data to network characteristics 6
Chair for Network Architectures and Services Technical University of Munich (TUM) Implementation ◮ Create a longest prefix match lookup tree of entries of the BGP data (libbgpdump, pytricia) ◮ Convert the TTL values to hop counts ◮ Assign an AS path to each entry in the capture file ◮ Compute the mean, variance, standard deviation of the hop counts and saving these intermediate results Till Wickenheiser – Correlating TTL data to network characteristics 6
Chair for Network Architectures and Services Technical University of Munich (TUM) Implementation ◮ Create a longest prefix match lookup tree of entries of the BGP data (libbgpdump, pytricia) ◮ Convert the TTL values to hop counts ◮ Assign an AS path to each entry in the capture file ◮ Compute the mean, variance, standard deviation of the hop counts and saving these intermediate results ◮ Create graphs using the matplotlib and seaborn libraries in an Ipython notebook Till Wickenheiser – Correlating TTL data to network characteristics 6
Chair for Network Architectures and Services Technical University of Munich (TUM) Runtime Data Number of entires File size Runtime IPv4 85 000 000 5 GB ∼ 8 h IPv6 2 000 000 200 MB ∼ 15 min Utilizing 16 CPUs running at 2.40GHz and 24GB of RAM. Till Wickenheiser – Correlating TTL data to network characteristics 7
Chair for Network Architectures and Services Technical University of Munich (TUM) Resulting graphs for mean hop count values Linear regression parameters Coefficient IPv4 IPv6 Slope 1.1448 1.8173 Intercept 7.7438 2.1843 R-squared 0.0522 0.1544 1.4 ∗ 10 − 9 P value 0 Standard error 0.0099 0.2873 Till Wickenheiser – Correlating TTL data to network characteristics 8
Chair for Network Architectures and Services Technical University of Munich (TUM) Resulting graphs Till Wickenheiser – Correlating TTL data to network characteristics 9
Chair for Network Architectures and Services Technical University of Munich (TUM) Resulting graphs for mean hop count values Till Wickenheiser – Correlating TTL data to network characteristics 10
Chair for Network Architectures and Services Technical University of Munich (TUM) Resulting graphs for mean hop count values Till Wickenheiser – Correlating TTL data to network characteristics 11
Chair for Network Architectures and Services Technical University of Munich (TUM) Resulting graphs for mean hop count values Linear regression parameters Coefficient IPv4 IPv6 Slope 1.1448 1.8173 Intercept 7.7438 2.1843 R-squared 0.0522 0.1544 1.4 ∗ 10 − 9 P value 0 Standard error 0.0099 0.2873 Till Wickenheiser – Correlating TTL data to network characteristics 12
Chair for Network Architectures and Services Technical University of Munich (TUM) Resulting graphs for mean hop count values Till Wickenheiser – Correlating TTL data to network characteristics 13
Chair for Network Architectures and Services Technical University of Munich (TUM) Resulting graphs for mean hop count values Till Wickenheiser – Correlating TTL data to network characteristics 14
Chair for Network Architectures and Services Technical University of Munich (TUM) Resulting graphs for mean hop count values Till Wickenheiser – Correlating TTL data to network characteristics 15
Chair for Network Architectures and Services Technical University of Munich (TUM) Minima per AS path value Till Wickenheiser – Correlating TTL data to network characteristics 16
Chair for Network Architectures and Services Technical University of Munich (TUM) Minima per AS path value Till Wickenheiser – Correlating TTL data to network characteristics 17
Chair for Network Architectures and Services Technical University of Munich (TUM) Outliers There are 7000 occurrences (2.5%) of the minimum hop count being below the AS path length. These only occurred during the investigations regarding the IPv4 data. Till Wickenheiser – Correlating TTL data to network characteristics 18
Chair for Network Architectures and Services Technical University of Munich (TUM) Outliers There are 7000 occurrences (2.5%) of the minimum hop count being below the AS path length. These only occurred during the investigations regarding the IPv4 data. Reasons for these exceptions could be: ◮ Wrongfully chosen initial TTL value Till Wickenheiser – Correlating TTL data to network characteristics 18
Chair for Network Architectures and Services Technical University of Munich (TUM) Outliers There are 7000 occurrences (2.5%) of the minimum hop count being below the AS path length. These only occurred during the investigations regarding the IPv4 data. Reasons for these exceptions could be: ◮ Wrongfully chosen initial TTL value ◮ Used BGP data does not represent a good model for the real routing path Till Wickenheiser – Correlating TTL data to network characteristics 18
Recommend
More recommend
