measuring and visualizing inter-domain routing dynamics with bgpath
TRANSCRIPT
![Page 1: Measuring and Visualizing Inter-domain Routing Dynamics with BGPATH](https://reader036.vdocument.in/reader036/viewer/2022071601/613d4db7736caf36b75bbb70/html5/thumbnails/1.jpg)
UNIVERSITÀ DEGLI STUDI ROMA TRE Dipartimento di Informatica e Automazione
Measuring and Visualizing Inter-domain Routing
Dynamics with BGPATH Luca Cittadini Tiziana Refice
Alessio Campisano Giuseppe Di Battista
Claudio Sasso
IEEE Symposium on Computers and Communications (ISCC 2008)
Jul 8th, 2008
![Page 2: Measuring and Visualizing Inter-domain Routing Dynamics with BGPATH](https://reader036.vdocument.in/reader036/viewer/2022071601/613d4db7736caf36b75bbb70/html5/thumbnails/2.jpg)
Outline
Motivations Goals Basic ideas Analyze a route change using BGPath Stream-based computation Conclusions and future work
![Page 3: Measuring and Visualizing Inter-domain Routing Dynamics with BGPATH](https://reader036.vdocument.in/reader036/viewer/2022071601/613d4db7736caf36b75bbb70/html5/thumbnails/3.jpg)
Motivations
Policy routing Flexibility Fault prone, difficult to debug, …
Monitoring and data collection Distributed and publicly available monitors Historical repositories (RIS, RV) Huge amount of data
Semi-automatic analysis tools Internet-wide (LinkRank, [NOMS’04]) ISP-oriented (BGPlay, [GD’03], Cyclops [SIGCOMM’06])
![Page 4: Measuring and Visualizing Inter-domain Routing Dynamics with BGPATH](https://reader036.vdocument.in/reader036/viewer/2022071601/613d4db7736caf36b75bbb70/html5/thumbnails/4.jpg)
Goals
Provide the user with valuable information about inter-domain routing dynamics, in order to help detection and analysis of network events Define metrics to describe inter-domain routing
behavior Compute metric on huge amount of input BGP data,
fulfilling near real time constrains Visualize computed values to easily detect inter-
domain routing events
Bridge the gap between Internet-wide and ISP-oriented approaches
![Page 5: Measuring and Visualizing Inter-domain Routing Dynamics with BGPATH](https://reader036.vdocument.in/reader036/viewer/2022071601/613d4db7736caf36b75bbb70/html5/thumbnails/5.jpg)
Basic ideas In theory… [NOMS’08]
Collect data from many monitors Evaluate data source reliability, in order to
filter out data collection ‘noise’ Model inter-domain routing dynamics in
terms of ‘usage’ of inter-domain links over time for all the routed prefixes
Analyze inter-domain routing of user-defined prefix, in order to identify network events affecting it
![Page 6: Measuring and Visualizing Inter-domain Routing Dynamics with BGPATH](https://reader036.vdocument.in/reader036/viewer/2022071601/613d4db7736caf36b75bbb70/html5/thumbnails/6.jpg)
Basic ideas … in practice [ISCC’08]
Efficient algorithms and implementations to compute data source reliability and link usage Huge amount of data (BGP updates)
• ~2 GB/day, ~50M updates/day • ~100K inter-domain links, ~250K prefixes,
~800 monitors Need for nearly real-time information, to detect and
analyze inter-domain routing outages
Effective visualization of inter-domain information (raw and computed) Multidimensional data (time, topology, monitors,
prefixes)
![Page 7: Measuring and Visualizing Inter-domain Routing Dynamics with BGPATH](https://reader036.vdocument.in/reader036/viewer/2022071601/613d4db7736caf36b75bbb70/html5/thumbnails/7.jpg)
Analyze a route change using BGPath
Use case: Fiber cuts in the Mediterranean Sea Outage period: 01/23-02/14, 2008
Let’s focus on SEA-ME-WE-4 cable cut as observed by AS17641 (Infotech, BD),
which originates 202.65.10.0/23
![Page 8: Measuring and Visualizing Inter-domain Routing Dynamics with BGPATH](https://reader036.vdocument.in/reader036/viewer/2022071601/613d4db7736caf36b75bbb70/html5/thumbnails/8.jpg)
Input: a route change
Launch BGPath on
the selected route change
BGPath integrates with BGPlay, in order to graphically select a BGP route change
Infotech
Infotech’s upstream
![Page 9: Measuring and Visualizing Inter-domain Routing Dynamics with BGPATH](https://reader036.vdocument.in/reader036/viewer/2022071601/613d4db7736caf36b75bbb70/html5/thumbnails/9.jpg)
BGPath - overview
input route
change
![Page 10: Measuring and Visualizing Inter-domain Routing Dynamics with BGPATH](https://reader036.vdocument.in/reader036/viewer/2022071601/613d4db7736caf36b75bbb70/html5/thumbnails/10.jpg)
Customer-provider hierarchy
Tier-1
Infotech
TELIANET
Place the route change in the Internet hierarchy, in order to assess the impact of the change on the network
CNH
old path
new path
AS17469 changes upstream
![Page 11: Measuring and Visualizing Inter-domain Routing Dynamics with BGPATH](https://reader036.vdocument.in/reader036/viewer/2022071601/613d4db7736caf36b75bbb70/html5/thumbnails/11.jpg)
BGPath - overview
![Page 12: Measuring and Visualizing Inter-domain Routing Dynamics with BGPATH](https://reader036.vdocument.in/reader036/viewer/2022071601/613d4db7736caf36b75bbb70/html5/thumbnails/12.jpg)
Monitor resets
Infer resets of connections between any monitor and the collection system, in order to identify unreliable data sources
All the monitors with routes towards
202.65.10.0/23
![Page 13: Measuring and Visualizing Inter-domain Routing Dynamics with BGPATH](https://reader036.vdocument.in/reader036/viewer/2022071601/613d4db7736caf36b75bbb70/html5/thumbnails/13.jpg)
BGPath - overview
![Page 14: Measuring and Visualizing Inter-domain Routing Dynamics with BGPATH](https://reader036.vdocument.in/reader036/viewer/2022071601/613d4db7736caf36b75bbb70/html5/thumbnails/14.jpg)
AS-path evolution
Plot the AS paths of prefixes over time, in order to help identify patterns in routing dynamics
prefixes originated by Infotech
AS-paths from Infotech to CNH
AS17469 changes upstreams AS24122 disconnected from Infotech
![Page 15: Measuring and Visualizing Inter-domain Routing Dynamics with BGPATH](https://reader036.vdocument.in/reader036/viewer/2022071601/613d4db7736caf36b75bbb70/html5/thumbnails/15.jpg)
AS-path evolution in the wild
![Page 16: Measuring and Visualizing Inter-domain Routing Dynamics with BGPATH](https://reader036.vdocument.in/reader036/viewer/2022071601/613d4db7736caf36b75bbb70/html5/thumbnails/16.jpg)
BGPath - overview
![Page 17: Measuring and Visualizing Inter-domain Routing Dynamics with BGPATH](https://reader036.vdocument.in/reader036/viewer/2022071601/613d4db7736caf36b75bbb70/html5/thumbnails/17.jpg)
‘Usage’ of inter-domain links
two perspectives: local : one monitor global : all monitors
Measure the # prefixes through inter-domain links, in order to pinpoint macro events affecting physical and/or logical topology
No prefixes use AS-link 701-17494
![Page 18: Measuring and Visualizing Inter-domain Routing Dynamics with BGPATH](https://reader036.vdocument.in/reader036/viewer/2022071601/613d4db7736caf36b75bbb70/html5/thumbnails/18.jpg)
‘Usage’ of inter-domain links in the wild
![Page 19: Measuring and Visualizing Inter-domain Routing Dynamics with BGPATH](https://reader036.vdocument.in/reader036/viewer/2022071601/613d4db7736caf36b75bbb70/html5/thumbnails/19.jpg)
Stream-based approach
![Page 20: Measuring and Visualizing Inter-domain Routing Dynamics with BGPATH](https://reader036.vdocument.in/reader036/viewer/2022071601/613d4db7736caf36b75bbb70/html5/thumbnails/20.jpg)
Time and space complexity
time: O(np)
space: O(p)
time: O(nl)
space: O(ep)
n: # route changes (~50M/d) p: # all known prefixes (250k)
l: max AS-path length (~7) e: # edges (~100k)
![Page 21: Measuring and Visualizing Inter-domain Routing Dynamics with BGPATH](https://reader036.vdocument.in/reader036/viewer/2022071601/613d4db7736caf36b75bbb70/html5/thumbnails/21.jpg)
Time and space performance
~1 min
~1.2 Gb (peak)
~1 min
15 min chunks
Overall, less than 3 min
Nearly real-time computation
~2.7 Gb (peak)
![Page 22: Measuring and Visualizing Inter-domain Routing Dynamics with BGPATH](https://reader036.vdocument.in/reader036/viewer/2022071601/613d4db7736caf36b75bbb70/html5/thumbnails/22.jpg)
Conclusions
BGPath, publicly available system with Global and ISP view of routing dynamics Efficient computation of data source reliability
and inter-domain link usage Effective graphical visualization of fine grained
and aggregated information about user-specified route changes
Already extensively used to analyze Mediterranean fiber cuts (Jan 2008) Hijacking of the YouTube prefix by Pakistan Telekom
(Feb 24th 2008)
![Page 23: Measuring and Visualizing Inter-domain Routing Dynamics with BGPATH](https://reader036.vdocument.in/reader036/viewer/2022071601/613d4db7736caf36b75bbb70/html5/thumbnails/23.jpg)
Future work
Extend BGPath’s user interface with additional routing information
Define new metrics, in order to better model inter-domain routing dynamics
Deploy pattern recognition modules, in order to automatically pinpoint network events and their root causes
Further BGP forensic analyses