1 / 34

Extracting Intra-Domain Topology from mrinfo Probing

Extracting Intra-Domain Topology from mrinfo Probing. JJ Pansiot (Lsiit, UdS) B Donnet, P Mérindol, O Bonaventure (INL, Louvain). Agenda. Background Part 1: mrinfo Context Probing Methodology Dataset Part 2: Delimiting AS Borders Router-2-AS Mapping Evaluation

flint
Download Presentation

Extracting Intra-Domain Topology from mrinfo Probing

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Extracting Intra-Domain Topology from mrinfo Probing JJ Pansiot (Lsiit, UdS) B Donnet, P Mérindol, O Bonaventure (INL, Louvain)

  2. Agenda • Background • Part 1: mrinfo • Context • Probing Methodology • Dataset • Part 2: Delimiting AS Borders • Router-2-AS Mapping • Evaluation • Part 3: Intra-Domain Study • Importance of Layer-2 • Conclusion

  3. Background • Internet graph seen either as • a graph of routers • Constructed for example with traceroute • a graph of domains (AS) • Constructed for example from BGP tables (routeviews) • or both • To study intra-domain graphs • To study interconnection of domains • (our paper at IMC) • need a way to overlay the AS graph on top of the router graph

  4. Part 1: mrinfo • Graph of routers • Usually obtained from traceroute tools • Difficulty to map IP adresses to routers • Aliasing • Need many sources and targets • To increase exhaustivity in edge discovery • Many artifacts • Eg load balancers, changes in routing • => try another way to get information: mrinfo

  5. Probing • Topology discovery using mrinfo • Uses IGMP message • ASK_NEIGHBORS • NEIGHBORS_REPLY • Output • All (multicast) interfaces of a given router • All (multicast) neighbor routers • mrinfo applied recursively • mrinfo-rec

  6. switch Probing

  7. Data set • Four years of data • collected daily • May 1st, 2004 to December 31st, 2008 • Recursive probing scheme: mrinfo-rec • Publicly available • http://svnet.u-strasbg.fr/mrinfo/index.html • (raw data) • http://inl.info.ucl.ac.be/content/mrinfo • (refined data)

  8. Data set: routers

  9. Data set: interfaces

  10. Part 2: Delimiting AS Borders

  11. IP to AS • First step : IP to AS mapping • Obtained from routeviews data • Allows to get a mapping on a given day • Discard some rare cases • Such as MOA (Multiple Origin AS) • Unknown prefixes • AS identified by an ASN (AS Number)

  12. Example routeviews prefix Next Hop AS path * 130.79.0.0 217.75.96.60 0 0 16150 1239 5511 2200 2259 i * 209.161.175.4 0 14608 19029 3356 5511 2200 2259 i * 129.250.0.85 11 0 2914 5511 2200 2259 i * 129.250.0.11 18 0 2914 5511 2200 2259 i * 216.140.8.59 2741 0 6395 5511 2200 2259 i * 216.140.2.59 3800 0 6395 5511 2200 2259 i * 208.186.154.35 0 0 5650 1239 5511 2200 2259 i … * 8.7.83.0/24 217.75.96.60 0 0 16150 6939 5650 12284 32808 32808 i * 209.161.175.4 0 14608 19029 5650 12284 32808 32808 i * 216.140.2.59 0 0 6395 5650 12284 32808 32808 i * 216.140.8.59 20 0 6395 5650 12284 32808 32808 i

  13. Router to AS Mapping • Problem • Graph of routers • Each router has • Many IP addresses (interfaces) • IP => prefix => ASN • IP addresses may belong to different ASes • For each router decide of the « correct » AS

  14. AS2 Or this? Router-2-AS Mapping This?

  15. Router-2-AS Mapping • Idea • Apply a series of rules • At each step a router r • Has an ASN, Asn(r) • An a confidence level c(r), 0 ≤ c(r) ≤ 1 • A rule applied at a given step, if c(r) < 1 • Increase the confidence level • And possibly changes ASN(r) • While remaining consistent with ASN already assigned • Once c(r) = 1, ASN(r) unchanged • => most reliable rules first

  16. Main rule : election elec • The first (main) rule is an election (elec) • router r is given the most frequent AS among its addresses (with tie breaker), ASN1 • Plus a confidence level c(r) • From 1 (all addresses in the same AS) • To 0 (same number of addresses in 2 distinct AS) c(r) = (frequency(ASN1)-frequency(ASN2)) / frequency(ASN1)

  17. Rule loopback lb • If a router has a loopback interface lb • As deduced from the dns name • Eg loop0.ar3.LON2.gblx.net • Then ASN(r) = ASN(lb) with c(r) =1

  18. Rule neighbor N • N rule allows to propagate « ASN certainty » • Assumption • On a P2P link the prefix of the link belongs to the AS of 1 of its end routers • If R1 is in AS1 (with c(R1) = 1) and link R1 - R2 is in AS2 • then ASN(R2) = AS2 with c(R2) =1 • Warning : not applied to multipoint links • Eg IXP

  19. Rule lan • Idea : if router r has an interface to a leaf Lan in ASN (leaf) • Then ASN(r) = ASN(leaf)

  20. Increase confidence and propagate • For remaining routers with 0 < c(r) < 1 • By decreasing order of c(r) • set c(r) to 1 • Propagate to neighbors (rule N)

  21. AS1 AS2 R1 R3 R2 AS1 AS1 AS2 C(R3)=0 AS2 customer of AS1 Rule c2p • Customer to provider rule c2p • Assumption: the link between a provider AS and its customer AS is usually taken in the provider address space • Need to know the AS relationship • Obtained from work by Caida => ASN(R3) = AS2 AS2 C(R3) = 1

  22. Evaluation • Subset of original dataset • 1 probing dataset/month • 56 samples • Purpose • Algorithm efficiency

  23. The power of election

  24. 90% solved after elec 95% solved after lan 0.45% unsolved at the end Other rules : refinement

  25. Closer look at steps

  26. Part 3: Preliminary topology analysis: Some results

  27. Global switch/router ratio

  28. AS3356 Level3

  29. AS1239 Sprint

  30. Les CRS ne répondent plus

  31. Conclusion • New tool for collecting topological data • mrinfo-rec • Intra-Domain topology delimitation • Importance of layer-2

  32. Conclusion • Jean-Jacques Pansiot, Pascal Mérindol, Benoit Donnet, Olivier Bonaventure.Extracting Intra-Domain Topology frommrinfoProbing. In Proc. PAM 2010, Zurich. • Pascal Mérindol, Virginie Van den Schrieck, Benoit Donnet, Olivier Bonaventure, Jean-Jacques Pansiot. Quantifying ASes Multiconnectivity using Multicast Information. In Proc. IMC 2009. • Preliminary analysis: • Jean-Jacques Pansiot. Local and Dynamic Analysis of Internet Multicast Router Topology. Annals of telecommunications, 2007. • Data publicly available • http://svnet.u-strasbg.fr/mrinfo/index.html (raw data) • http://inl.info.ucl.ac.be/content/mrinfo (refined data)

  33. Output of mrinfo 128.103.15.25 (bdrgw2-vl-15-core.net.harvard.edu) [version 12.2]: 192.5.66.204 -> 0.0.0.0 (local) [1/0/pim/querier/leaf] 128.103.15.25 -> 128.103.15.23 (hlkgw1-vl-15-core.net.harvard.edu) [1/0/pim] 128.103.15.25 -> 128.103.15.30 (perkgw1-vl-15-core.net.harvard.edu) [1/0/pim] 128.103.15.25 -> 128.103.15.26 (war10gw1-vl-15-core.net.harvard.edu) [1/0/pim] 128.103.15.25 -> 128.103.15.21 (oxgw3-vl-15-core.net.harvard.edu) [1/0/pim] 128.103.15.25 -> 128.103.15.16 (arsgw1-vl-15-core.net.harvard.edu) [1/0/pim] 128.103.15.25 -> 128.103.15.34 (bdrgw1-ge-3-1-core.net.harvard.edu) [1/0/pim] 128.103.15.25 -> 128.103.15.22 (lowgw1-vl-15-core.net.harvard.edu) [1/0/pim] 128.103.15.25 -> 128.103.15.15 (oxgw2-vl-15-core.net.harvard.edu) [1/0/pim] 128.103.15.25 -> 128.103.15.20 (nh175gw1-vl-15-core.net.harvard.edu) [1/0/pim] 128.103.15.25 -> 128.103.15.27 (st8gw1-vl-15-core.net.harvard.edu) [1/0/pim] 128.103.17.25 -> 128.103.17.23 (hlkgw1-vl-17-rcore.net.harvard.edu) [1/0/pim] 128.103.17.25 -> 128.103.17.30 (perkgw1-vl-17-rcore.net.harvard.edu) [1/0/pim] 128.103.17.25 -> 128.103.17.26 (war10gw1-vl-17-rcore.net.harvard.edu) [1/0/pim] 128.103.17.25 -> 128.103.17.21 (oxgw3-vl-17-rcore.net.harvard.edu) [1/0/pim] 128.103.17.25 -> 128.103.17.17 (cftgw1-vl-17-rcore.net.harvard.edu) [1/0/pim] 128.103.17.25 -> 128.103.17.16 (arsgw1-vl-17-rcore.net.harvard.edu) [1/0/pim] 128.103.17.25 -> 128.103.17.34 (bdrgw1-ge-4-1-rcore.net.harvard.edu) [1/0/pim] 128.103.17.25 -> 128.103.17.27 (st8gw1-vl-17-rcore.net.harvard.edu) [1/0/pim] 128.103.17.25 -> 128.103.17.22 (lowgw1-vl-17-rcore.net.harvard.edu) [1/0/pim] 128.103.17.25 -> 128.103.17.20 (nh175gw1-vl-17-rcore.net.harvard.edu) [1/0/pim] 128.103.252.145 -> 128.103.252.146 (lmagw1-vl-100-core.net.harvard.edu) [1/0/pim] 192.5.89.6 -> 192.5.89.5 (ATM10-400-OC12-GIGAPOPNE.nox.org) [1/0/pim/querier]

  34. After processing ID7 ROUTEUR 128.103.15.25, AS1742 INTER bdrgw2-vl-15-core.net.harvard.edu LOCAL 192.5.66.204(AS1742) -> 0.0.0.0 SWITCH 128.103.15.25(AS1742) -> (AS1742) ID10001 AS1742 ( bdrgw2-vl-15-core.net.harvard.edu -> ) SWITCH 128.103.17.25(AS1742) -> (AS1742) ID9656 AS1742 ( bdrgw2-vl-17-rcore.net.harvard.edu -> ) GLOBAL 128.103.252.145(AS1742) -> 128.103.252.146 (AS1742) ID13 AS1742 ( bdrgw2-vl-100-core.net.harvard.edu -> lmagw1-vl-100-core.net.harvard.edu) GLOBAL 192.5.89.6(AS10578) -> 192.5.89.5 (AS10578) ID3827 AS10578 ( HARVARD-GIGAPOPNE.nox.org -> ATM10-400-OC12-GIGAPOPNE.nox.org) ID10001 SWITCH 128.103.15.0/26, AS1742 INTRA () -> 128.103.15.16 (AS1742) ID1 AS1742 ( -> arsgw1-vl-15-core.net.harvard.edu) () -> 128.103.15.21 (AS1742) ID4 AS1742 ( -> oxgw3-vl-15-core.net.harvard.edu) () -> 128.103.15.20 (AS1742) ID3 AS1742 ( -> nh175gw1-vl-15-core.net.harvard.edu) () -> 128.103.15.30 (AS1742) ID10 AS1742 ( -> perkgw1-vl-15-core.net.harvard.edu) () -> 128.103.15.25 (AS1742) ID7 AS1742 ( -> bdrgw2-vl-15-core.net.harvard.edu) () -> 128.103.15.34 (AS1742) ID11 AS1742 ( -> bdrgw1-ge-3-1-core.net.harvard.edu) () -> 128.103.15.23 (AS1742) ID6 AS1742 ( -> hlkgw1-vl-15-core.net.harvard.edu) () -> 128.103.15.27 (AS1742) ID9 AS1742 ( -> st8gw1-vl-15-core.net.harvard.edu) () -> 128.103.15.15 (AS1742) ID0 AS1742 ( -> oxgw2-vl-15-core.net.harvard.edu) () -> 128.103.15.22 (AS1742) ID5 AS1742 ( -> lowgw1-vl-15-core.net.harvard.edu) () -> 128.103.15.26 (AS1742) ID8 AS1742 ( -> war10gw1-vl-15-core.net.harvard.edu)

More Related