1. ETH Zurich – Distributed Computing – www.disco.ethz.ch On Consistent Migration of Flows in SDNs Sebastian Brandt, Klaus-Tycho Förster, Roger Wattenhofer April 12, 2016 @ INFOCOM 2016 – San Francisco

2. The Internet General Idea: Separate data & control plane in a network Centralized controller updates networks rules for optimization – Controller (control plane) updates the switches/routers (data plane)

3. Central Control? General Idea: Separate data & control plane in a network Centralized controller updates networks rules for optimization – Controller (control plane) updates the switches/routers (data plane)

4. Think: Google, Amazon, Microsoft Own WAN = Expensive

5. Software Defined Networking (SDN) +

6. Network Updates old network rules new network rules network updates

7. Network Updates old network rules new network rules network updates State of the art: (Partial) moves of flows using linear programming (LPs), e.g., SWAN [Hong et al., SIGCOMM 2013], zUPDATE [Liu et al., SIGCOMM 2013] Dionysus [Jin et al., SIGCOMM 2014]

8. Network Updates old network rules new network rules network updates Open problems: When are network updates in a consistent manner possible? How can we decide fast?

9. Network Updates old network rules new network rules network updates This paper: Addresses the case of splittable multi-commodity flows

10. A Small Sample Network

11. Green wants to send as well

12. Congestion!

13. This would work

14. So lets go back

15. But Red is a bit Slow..

16. Congestion Again!

17. Appears in Practice “Data plane updates may fall behind the control plane acknowledgments and may be even reordered.” Kuzniar et al., PAM 2015 “some switches can ‘straggle,’ taking substantially more time than average (e.g., 10-100x) to apply an update” Jin et al., SIGCOMM 2014 “…the inbound latency is quite variable with a […] standard deviation of 31.34ms…” He et al., SOSR 2015

18. So lets go Back …

19. First, Red switches

20. Then, Blue …

21. And then, Green …

22. Done

23. Consistent Migration of Flows

26. Update 1 of 2

27. Consistent Migration of Flows Update 1 of 2

28. Consistent Migration of Flows Update 2 of 2

29. Consistent Migration of Flows No slack on flow edges?

30. Consistent Migration of Flows Alternate routes?

31. Consistent Migration of Flows

34. SWAN: LP-approach with binary search 1 update? 2 updates? 4 updates? …

35. Consistent Migration of Flows SWAN: LP-approach with binary search 1 update? 2 updates? 4 updates? …

36. Consistent Migration of Flows SWAN: LP-approach with binary search

37. Consistent Migration of Flows Open problem: Can we decide in (polynomial) time? “Halting Problem” Flow migration yes LP no

38. Overview of the Remaining Talk 1.Yes, we can (decide in polynomial time) 2.What to do if we cannot migrate consistently? 3.Last: NP-hardness for unsplittable flows

39. To Slack or not to Slack?

40. What if not? Try to create slack

41. To Slack or not to Slack? Combinatorial approach Augmenting paths

42. Combinatorial Approach Move single commodities at a time u v

43. Combinatorial Approach Where to increase flow? u v

44. Combinatorial Approach Where to push back flow? u v

45. Combinatorial Approach Resulting residual network u v

46. Combinatorial Approach u v

47. High-level Algorithm Idea

48. Overview of the Remaining Talk 1.Yes, we can (decide in polynomial time) 2.What to do if we cannot migrate consistently? 3.Last: NP-hardness for unsplittable flows

49. What to do if we cannot Migrate Consistently? Option 1: Migrate, but reduce congestion B4: Optimize (Jain et al., SIGCOMM 13) Dionysus: Rate-limit some flows (Jin et al., SIGCOMM 14) Time-based updates (E.g., Mizrahi et al., INFOCOM 15/16) …

50. What to do if we cannot Migrate Consistently? Option 1: Migrate, but reduce congestion B4: Optimize (Jain et al., SIGCOMM 13) Dionysus: Rate-limit some flows (Jin et al., SIGCOMM 14) Time-based updates (E.g., Mizrahi et al., INFOCOM 15/16) … Wed, 08:30: Grand Ballroom B

51. Option 2: Increase Demands Consistently Idea: Don’t change where no slack is possible

52. Overview of the Remaining Talk 1.Yes, we can (decide in polynomial time) 2.What to do if we cannot migrate consistently? 3.Last: NP-hardness for unsplittable flows

53. NP-Hardness for Unsplittable Flows

54. Summary Consistent migration of flows Decidable in polynomial time No consistent migration possible? Can use LP to maximize demands Unsplittable flow migration NP-hard

55. ETH Zurich – Distributed Computing – www.disco.ethz.ch On Consistent Migration of Flows in SDNs Sebastian Brandt, Klaus-Tycho Förster, Roger Wattenhofer April 12, 2016 @ INFOCOM 2016 – San Francisco