routing without flow control
play

Routing without Flow Control Costas Busch Rensselaer Polytechnic - PowerPoint PPT Presentation

Routing without Flow Control Costas Busch Rensselaer Polytechnic Institute Maurice Herlihy Brown University Roger Wattenhofer Microsoft Research 1 The network: n mesh n n n Discrete time Bi-directional links At most one


  1. Routing without Flow Control Costas Busch Rensselaer Polytechnic Institute Maurice Herlihy Brown University Roger Wattenhofer Microsoft Research 1

  2. The network: n  mesh n n n • Discrete time • Bi-directional links • At most one packet per link direction 2

  3. Dynamic Routing: Packets are injected continuously destination 3

  4. A new packet can be injected when there is a free link : A link direction is empty 4

  5. Most dynamic routing algorithms use flow control : Don’t utilize all the free links Disadvantage: Network is under-utilized 5

  6. Our Routing Algorithm: • No flow control • Utilizes all the free links Advantage: Network is fully-utilized 6

  7. Features of our algorithm: • Dynamic • Hot potato • Optimal delivery time: ( n ) O • Injection time guaranty: ( n ) O 7

  8. Talk Outline The Algorithm Time Analysis Stability Future Work 8

  9. Hot-Potato Routing: • Nodes are buffer-less • Packets are immediately forwarded 9

  10. Conflicts 10

  11. Conflicts Conflict 11

  12. Conflicts Deflected 12

  13. Packet states: Priorities: Running high Excited Active Sleeping low 13

  14. Sleeping packet destination Random destination 14

  15. Sleeping packet destination Follows a path to destination 15

  16. Sleeping packet becomes Active with n probability   1  n     n 16

  17. Active packet Follows a greedy path 17

  18. Active packet Follows a greedy path 18

  19. Active packet A conflict situation 19

  20. Active packet Conflict A conflict situation 20

  21. Active packet Deflected A conflict situation 21

  22. Active packet becomes Deflected Excited with probability   1     p   n A conflict situation 22

  23. Excited packet Follows a one-bend path 23

  24. Excited packet becomes Running Follows a one-bend path 24

  25. Running packet Follows a one-bend path 25

  26. Talk Outline The Algorithm Time Analysis Stability Future Work 26

  27. Good condition for a column: at most non-sleeping packets 10 n with destination in the column 27

  28. Expected delivery time for one packet:    n (when the destination column is in good condition) 28

  29. Initially a packet is sleeping    In expected time steps n becomes active We will show: An active packet is delivered   in expected time steps O n 29

  30. Interrupting a one-bend path Excited Time 1 30

  31. Interrupting a one-bend path Running Time 2 31

  32. Interrupting a one-bend path Excited Running Time 2 32

  33. Interrupting a one-bend path Running Running Time 3 33

  34. Interrupting a one-bend path Running conflict Running Time 4 34

  35. Interrupting a one-bend path deflected Active Running Time 5 35

  36. No interruption probability: (  m 1 p ) Excitement probability Number of non-sleeping packets with destinations in same column 36

  37. No interruption probability: p m   ) ( 1 c   1  n constant        n (when the destination column is in good condition) 37

  38. Probability of success after a deflection:   1      p c   n Expected number of   deflections until success:  n Expected delivery time   for an active packet: O n 38

  39. Talk Outline The Algorithm Time Analysis Stability Future Work 39

  40. Divide time in time periods: t 6 n Examine the condition of a column 40

  41. 1 time period e   n 1 e  n Good condition Bad condition   10 m n 10 m n 41

  42. 1 time period e   n 1 e  n Good condition Bad condition   ne  10 m n  n 10 m n 1 ne  n 4n time periods 42

  43. 1 time period e   n 1 e  n Good condition Bad condition   10 m n 10 m n 43

  44. Proof Outline In a time period: • At most new non-sleeping 2 n packets are generated with destinations in the column • At least non-sleeping packets 2 n are delivered (if )  8 m n 44

  45. Good condition Bad condition   ne  10 m n  n 10 m n 1 ne  n 4n time periods 45

  46. Proof Outline In a time period: • At most new non-sleeping 2 n packets are generated with destinations in the column • At least non-sleeping packets 3 n are delivered 46

  47. Consequences: • Most of the time, the columns are in good condition • Each packet is delivered in   expected time  n 47

  48. Talk Outline The Algorithm Time Analysis Stability Future Work 48

  49. • Arbitrary network topologies • De-randomization: Determistic destinations No randomized algorithm • Small number of packets 49

Recommend


More recommend