1. Chapter 4 NETWORK LAYER Computer Networks Summer 2007 Distributed Computing Group Dynamic PowerPoint Slides Only

2. Distributed Computing Group Computer Networks R. Wattenhofer4/2 Remember: Count to Infinity Problem bac c: 2c: 1 c: 3 c: 4 c: 5 c: 6 c: 7 c: 8

3. Distributed Computing Group Computer Networks R. Wattenhofer4/3 BGP does not count to infinity DestinationDirDstPath Zurichc4cdeZ bacde DestinationDirDstPath Zurichb5bcdeZ Zurich

4. Distributed Computing Group Computer Networks R. Wattenhofer4/4 DestinationDirDstPath Zurichc4cdeZ bacde DestinationDirDstPath Zurichb5bcdeZ Zurich withdraw Zurich BGP does not count to infinity

5. Distributed Computing Group Computer Networks R. Wattenhofer4/5 BGP Basics Continued DestinationDirDstPath Zurichc4cdeZ bacde DestinationDirDstPath Zurichb5bcdeZ Zurich announce bcdeZ

6. Distributed Computing Group Computer Networks R. Wattenhofer4/6 BGP Basics Continued DestinationDirDstPath Zurichc4cdeZ Zurichf3feZ bacde DestinationDirDstPath Zurichb4bfeZ Zurich announce bfeZ f active backup 30s

7. Distributed Computing Group Computer Networks R. Wattenhofer4/7 BGP Basics Continued DestinationDirDstPath Zurichc4cdeZ Zurichf3feZ bacde DestinationDirDstPath Zurichb4bfeZ Zurich announce bcdeZ f active backup

8. Distributed Computing Group Computer Networks R. Wattenhofer4/8 BGP (Border Gateway Protocol) BGP is the Internet de-facto standard Path Vector protocol 1)Receive BGP update (announce or withdrawal) from a neighbor. 2)Update routing table. 3)Does update affect active route? (Loop detection, policy, etc.) If yes, send update to all neighbors that are allowed by policy. MinRouteAdver: At most 1 announce per neighbor per 30+jitter seconds. Store the active routes of the neighbors.

9. Distributed Computing Group Computer Networks R. Wattenhofer4/9 Internet Architecture DestinationDirDstPath Zurichc4cdeZ 172.30.160/19R141239 1 3561 iBGP Route flap dampening Multipath Soft configuration …

10. Distributed Computing Group Computer Networks R. Wattenhofer4/10 Internet inter-AS routing: BGP BGP messages exchanged using TCP. BGP messages –OPEN: opens TCP connection to peer and authenticates sender –UPDATE: advertises new path (or withdraws old) –KEEPALIVE keeps connection alive in absence of UPDATES; also ACKs OPEN request –NOTIFICATION: reports errors in previous msg; also used to close connection Policy –Even if two BGP routers are connected they may not announce all their routes or use all the routes of the other –Example: if AS A does not want to route traffic of AS B, then A should simply not announce anything to B.

11. Distributed Computing Group Computer Networks R. Wattenhofer4/11 Robustness of BGP We are interested in routes to destination d. Nodes a,b,c all have the policy to prefer a 2-hop route through their clockwise neighbor over a direct 1-hop route to destination d. bc d a

12. Distributed Computing Group Computer Networks R. Wattenhofer4/12 BGP Update Traffic (Mae-East) Cisco bug withdraw loop is fixed with IOS release.

13. Distributed Computing Group Computer Networks R. Wattenhofer4/13 NSFNet Backbone Regional Campus Hello/EGP Internet Evolution: NSFNet (1995)

14. Distributed Computing Group Computer Networks R. Wattenhofer4/14 AS1 AS2 AS3 AS4 AS8 AS5 AS7 AS6 BGP Internet Evolution: Today

15. Distributed Computing Group Computer Networks R. Wattenhofer4/15 Experimental Setup Analyzed secondary paths of 20x20 AS pairs: –Inject and monitor BGP faults. –Survey providers on policies.

16. Distributed Computing Group Computer Networks R. Wattenhofer4/16 180 BGP Convergence Times

17. Distributed Computing Group Computer Networks R. Wattenhofer4/17 BGP Convergence Results If a link comes up, the convergence time is in the order of time to forward a message on the shortest path. If a link goes down, the convergence time is in the order of time to forward a message on the longest path.

18. Distributed Computing Group Computer Networks R. Wattenhofer4/18 a bcdef p a:p e:ap a:p d:ap a:p c:ap a:p b:ap a:p p:p Intuition for Slow Convergence

19. Distributed Computing Group Computer Networks R. Wattenhofer4/19 a bcdef p a:p e:ap a:p d:ap a:p c:ap a:p b:ap a:p p:p Os Intuition for Slow Convergence

20. Distributed Computing Group Computer Networks R. Wattenhofer4/20 a bcdef p a:p e:ap a:p d:ap a:p c:ap a:p b:ap a:p p:p WWWWW Intuition for Slow Convergence

21. Distributed Computing Group Computer Networks R. Wattenhofer4/21 de a:p c:ap W a c Intuition for Slow Convergence

22. Distributed Computing Group Computer Networks R. Wattenhofer4/22 de a:p c:ap dcap a c Intuition for Slow Convergence

23. Distributed Computing Group Computer Networks R. Wattenhofer4/23 a bcdef p a:p e:ap a:p d:ap a:p c:ap a:p b:ap a:p p:p O.1s Wdcapedapcbap Intuition for Slow Convergence

24. Distributed Computing Group Computer Networks R. Wattenhofer4/24 a bcdef p e:apd:apc:apb:ap- - O.2s dcapedapcbapW c:bap d:cape:dap Intuition for Slow Convergence

25. Distributed Computing Group Computer Networks R. Wattenhofer4/25 a bcdef p e:dapd:capc:bap-- - Wdcbapedcap 30s!!! Intuition for Slow Convergence

26. Distributed Computing Group Computer Networks R. Wattenhofer4/26 a bcdef p e:dcapd:cbap--- - Wedcbap 60s Intuition for Slow Convergence

27. Distributed Computing Group Computer Networks R. Wattenhofer4/27 a bcdef p e:dcbap---- - W 90s Intuition for Slow Convergence

28. Distributed Computing Group Computer Networks R. Wattenhofer4/28 a bcdef p Convergence in the time to forward a message on the longest path. Intuition for Slow Convergence

29. Distributed Computing Group Computer Networks R. Wattenhofer4/29 a p c f h g d e i b j The longest path…

30. Distributed Computing Group Computer Networks R. Wattenhofer4/30 a p c f h g d e i b j … is NP-complete (APX)

31. Distributed Computing Group Computer Networks R. Wattenhofer4/31 TimeBGP Message/Event 10:40:302129 withdraws p 10:41:082117 announces 5696 2129 p 10:41:322117 announces 1 5696 2129 p 10:41:502117 announces 2041 3508 3508 4540 7037 1239 5696 2129 p 10:42:172117 announces 1 2041 3508 3508 4540 7037 1239 5696 2129 p 10:43:052117 announces 2041 3508 3508 4540 7037 1239 6113 5696 2129 p 10:43:352117 announces 1 2041 3508 3508 4540 7037 1239 6113 5696 2129 p 10:43:592117 withdraws p Example of BGP Convergence

32. Distributed Computing Group Computer Networks R. Wattenhofer4/32 a bcdef p edap W edcap edcbap W Remember the Example

33. Distributed Computing Group Computer Networks R. Wattenhofer4/33 What might help? Idea: Attach a cause tag to the withdrawal message identifying the failed link/node (for a given prefix). It can be shown that a cause tag reduces the convergence time to the shortest path Problems –Since BGP is widely deployed, it cannot be changed easily –ISPs (ASs) dont like the world to know that it is their link that is not stable, and cause tags do exactly that. –Race conditions make the cause tags protocol intricate

34. Distributed Computing Group Computer Networks R. Wattenhofer4/34 a bcdef p a:p e:ap a:p d:ap a:p c:ap a:p b:ap a:p p:p Example with BGP-CT (Cause Tags)

35. Distributed Computing Group Computer Networks R. Wattenhofer4/35 a bcdef p a:p e:ap a:p d:ap a:p c:ap a:p b:ap a:p p:p Os Example with BGP-CT

36. Distributed Computing Group Computer Networks R. Wattenhofer4/36 a bcdef p a:p e:ap a:p d:ap a:p c:ap a:p b:ap a:p p:p W(ap) O.1s Example with BGP-CT

37. Distributed Computing Group Computer Networks R. Wattenhofer4/37 p bc x ef Convergence Time using Cause Tags

38. Distributed Computing Group Computer Networks R. Wattenhofer4/38 p bc x ef Convergence Time using Cause Tags

39. Distributed Computing Group Computer Networks R. Wattenhofer4/39 p bc x ef Convergence in the time to forward a message on the new shortest path (instead of the longest). Convergence Time using Cause Tags