1 / 19

IXPs: Mapped?

IXPs: Mapped?. Brice Augustin* Université Pierre et Marie Curie – Paris 6 (* Work done while at AT&T Labs – Research) Balachander Krishnamurthy and Walter Willinger AT&T Labs – Research. Introduction. Internet Exchange Points (IXPs) Physical infrastructures to peer and exchange traffic

berny
Télécharger la présentation

IXPs: Mapped?

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. IXPs: Mapped? Brice Augustin* Université Pierre et Marie Curie – Paris 6 (* Work done while at AT&T Labs – Research) Balachander Krishnamurthy and Walter Willinger AT&T Labs – Research

  2. Introduction • Internet Exchange Points (IXPs) • Physical infrastructures to peer and exchange traffic • IXPs play an important role … • Big IXPs carry petabytes/day • Costs savings by bypassing Tier-1s • … but little is known about them • IXP databases: manually maintained; unknown quality • Few studies [Xu 2004], [He 2007]

  3. Typical peering at an IXP Content Provider 1 Content Provider 2 AS2 AS1 AS3 layer-2 switch AS5 AS4

  4. Motivation • Goal: build IXP peering matrices • Get as close as possible to ground truth • One step towards getting IXP traffic matrices • Useful to: • Understand peering activities at IXPs • Role of IXPs in AS eco-system & associated economics • Discover additional peering links • Current AS maps miss lots of p2p links • These links might be at IXPs

  5. Assembling Datasets • Designed a methodology to identify IXPs: • Existence, Membership, Peering matrices • Mined existing datasets • Good starting point, but not enough • Collected new datasets based on • Systematic efforts to obtain most complete inputs • State-of-art detection techniques • Little room for improvement

  6. Analyzing Datasets • Compare datasets’ contributions • IXP existence, members, peerings • Validation of the discovered peering matrices • Assign confidence ranks to peerings • Analysis of membership and peerings at IXPs

  7. Identifying IXPs in traceroute data [He 2007] AS2 AS1 B A  hop 1 layer-2 switch E C D  AS4 AS3  hop 2 IP addr in IXP prefix hop 3

  8. Technique 1: Targeted traceroutes Traceroute to other IXP members Traceroute to other IXP members AS7 AS2 AS6 T2 AS8 T3 T1 AS3 AS1 IXP T1 Traceroute server AS4 (Found 2.3K such servers on traceroute.org, PeeringDB) AS5

  9. Technique 2: LSRR traceroutes Source-routed traceroutes from PlanetLab node Source-routing-capable router R1 AS2 (Found such a router at 847 out of 4K IXP members) AS1 IXP AS3 AS4 R1 AS5

  10. Technique 3: BGP summaries • List of a BGP router’s peering sessions • Remote AS number and IP address AS2 AS1 AS3 IXP AS5 L1 BGP Looking Glass L1 BGP sum at L1: Found 1.1K such LGs on traceroute.org, Routeviews,.. AS1 @IXP1 AS3 @IXP3 AS4 Peering session

  11. Pre-requisites • List of IXPs with prefixes • Source: IXP db, websites • Prefixes for 278 out of 352 IXPs • List of IXP members • Source: IXP db, websites, DNS names • 4K members for 275 out of 278 IXPs • Public measurement infrastructure • 2.3K Traceroute servers, 1.1K BGP LGs • AS map augmented with new peering links (~130K)

  12. Datasets summary • Traceroutes (Targeted, LSRR), BGP sums, Pings • Collected Feb-Apr 2009 • CAIDA • 26 traceroute servers • Analyzed 1 snapshot in Apr 2009 • DIMES • 17K agents running traceroutes • Analyzed Feb 2009 dataset • Planetlab • Traceroute from 254 nodes to all IXP members • Collected in Dec 2008

  13. # IXPs detected • Detected 223 out of 278 IXPs • Checked 55 undetected IXPs: • Discussions with IXP operators, network admins, PCH team • IXP websites, whois • Status: • 11 active, 6 undetectable • 22 defunct (MAEs, NAPs…) • 11 Planned/not IXP/down • 5 unknown

  14. Peerings at IXPs • Validated 44K out of 58K • Of those 44K, 18K are not in UCLA’s AS map

  15. Validation • Detecting IXP peerings is subject to errors • “Ground truth” for IXP existence and members • IXP databases and BGP summaries • Need a method to validate IXP peerings • Assign a “high” confidence to peerings if: • We observed the peering in both directions, • i.e. AS1  IXP  AS2 and AS2  IXP  AS1 • or both ASes belong to a valid list of IXP members • i.e. IXP members detected in BGP sums

  16. Peerings per IXP AMS-IX (Amsterdam) DE-CIX (Frankfurt) LINX (London) # of peerings IXPs ranked by # of peerings

  17. Members per IXP AMS-IX (Amsterdam) DE-CIX (Frankfurt) LINX (London) Any2, Equinix Ashburn JPNAP Tokyo # of members IXPs ranked by # of members

  18. Conclusion • Obtained data for most IXPs • 223 detected IXPs, 44K validated peerings • Validation is possible through • Ground truth for IXP existence and membership • Confidence ranking for peerings • Improvements will require • Better coverage (i.e. sources in good locations) • Additional datasets

  19. Data Available http://www-rp.lip6.fr/~augustin/ixp/

More Related