1. Diffusion Mechanisms for Active Queue Management Department of Electrical and Computer Engineering University of Delaware May 19th / 2004 Rafael Nunez nunez@ece.udel.edu Gonzalo Arce arce@ece.udel.edu

2. 2 Diffusion Mechanisms for Active Queue Management Introduction Diffusion Early Marking Performance Conclusions and Future Work

3. 3 The Internet Today

4. 4 Congestion Desirable control: distributed, simple, stable and fair.

5. 5 Problems with Tail Dropping Penalizes bursty traffic Discriminates against large propagation delay connections. Global synchronization.

6. 6 Active Queue Management (AQM) Random Early Detection (Floyd and Jacobson, 1993) Router becomes active in congestion control. RED has been deployed in some Cisco routers.

7. 7 Random Early Detection (RED) Random packet drops in queue. Drop probability based on average queue: Four parameters:  q min  q max  P max  w q (overparameterized)

8. 8 Queue Behavior in RED

9. 9 Queue Behavior in RED (2) 20 new flows every 20 seconds Wq = 0.01 Wq = 0.001

10. 10 Other AQM’s Schemes Adaptive RED, REM, GREEN, BLUE,… Problems:  Over-parameterization  Not easy to implement in routers  Not much better performance than drop tail

11. 11 Diffusion Mechanisms for Active Queue Management Introduction Diffusion Early Marking Performance Conclusions and Future Work √

12. 12 Diffusion Mechanisms for AQM Instantaneous queue size. Better packet marking strategy. Simplified parameters.

13. 13 Probability of Marking a Packet Gentle RED function closely follows: (A)

14. 14 Evolution of the Congestion Window TCP in steady state: (B)

15. 15 Traffic in the Network Congestion Window = Packets In The Pipe + Packets In The Queue Or: (C) From (A), (B), (C), and knowing that : where

16. 16 Probability Function

17. 17 Error Diffusion Packet marking is analogous to halftoning:  Convert a continuous gray-scale image into black or white dots  Packet marking reduces to quantization Error diffusion: The error between input (continuous) and output (discrete) is incorporated in subsequent outputs.

18. 18 Diffusion Mechanism

19. 19 Diffusion Mechanism

20. 20 Diffusion Mechanism

21. 21 Diffusion Mechanism

22. 22 Diffusion Mechanism

23. 23 Diffusion Mechanism

24. 24 Diffusion Mechanism

25. 25 Diffusion Mechanism

26. 26 Diffusion Mechanism

27. 27 Diffusion Mechanism

28. 28 Diffusion Mechanism

29. 29 Diffusion Mechanism

30. 30 Algorithm Summary Diffusion Early Marking decides whether to mark a packet or not as: Where: M=2, b 1 =2/3, b 2 =1/3 Remember:

31. 31 Diffusion Mechanisms for Active Queue Management Introduction Diffusion Early Marking Performance Conclusions and Future Work √ √

32. 32 Dropping Packets

33. 33 Window Size 2 Flows: 10 Flows

34. 34 Stability of the Queue 100 long lived connections (TCP/Reno, FTP) Desired queue size = 30 packets

35. 35 Changing the number of flows 20 new flows every 20 seconds

36. 36 Diffusion Mechanisms for Active Queue Management Introduction Diffusion Early Marking Performance Conclusions and Future Work √ √ √

37. 37 Conclusions and Future Work Queue length stabilized and controlled without adjusting parameters. Diffusion mechanism improves the behavior of the proposed AQM scheme. Future Work:  Estimation of parameters  Analyze more traffic scenarios  Compare with other AQMs  Use diffusion mechanism in other AQMs

38. 38 Diffusion Mechanisms for Active Queue Management Introduction Diffusion Early Marking Parameter Estimation Conclusions and Future Work √ √ √ √