Diffusion Mechanisms for Active Queue Management

1
Diffusion Mechanisms for Active Queue Management
Rafael Nunez nunez_at_ece.udel.edu
Gonzalo Arce arce_at_ece.udel.edu

• Department of Electrical and Computer Engineering
• University of Delaware
• May 19th / 2004

2
Diffusion Mechanisms for Active Queue Management

• Introduction
• Diffusion Early Marking
• Model Optimizations
• Parameter Estimation
• Performance
• Conclusions and Future Work

3
The Internet Today
4
Congestion

• Desirable control distributed, simple, stable
  and fair.

5
Problems with Tail Dropping

• Penalizes bursty traffic
• Discriminates against large propagation delay
  connections.
• Global synchronization.

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
Random Early Detection (RED)

• Random packet drops in queue.
• Drop probability based on average queue

• Four parameters
• qmin
• qmax
• Pmax
• wq
• (overparameterized)

8
Queue Behavior in RED
9
Queue Behavior in RED (2)

• 20 new flows every 20 seconds

• Wq 0.01

• Wq 0.001

10
Other AQMs Schemes

• Adaptive RED, REM, GREEN, BLUE,
• Problems
• Over-parameterization
• Not easy to implement in routers
• Not much better performance than drop tail

11
Diffusion Mechanisms for Active Queue Management

• Introduction
• Diffusion Early Marking
• Model Optimizations
• Parameter Estimation
• Performance
• Conclusions and Future Work

v
12
Diffusion Mechanisms for AQM

• Instantaneous queue size.
• Better packet marking strategy.
• Simplified parameters.

13
Probability of Marking a Packet

• Gentle RED function closely follows

(A)
14
Evolution of the Congestion Window

• TCP in steady state

(B)
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
Probability Function
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
Diffusion Mechanism
19
Diffusion Mechanism
20
Diffusion Mechanism
21
Diffusion Mechanism
22
Diffusion Mechanism
23
Diffusion Mechanism
24
Diffusion Mechanism
25
Diffusion Mechanism
26
Diffusion Mechanism
27
Diffusion Mechanism
28
Diffusion Mechanism
29
Diffusion Mechanism
30
Diffusion Mechanisms for Active Queue Management

• Introduction
• Diffusion Early Marking
• Model Optimizations
• Parameter Estimation
• Performance
• Conclusions and Future Work

v
v
31
Threshold Modulation

• Incorporate the queues rate of change in order
  to obtain faster response.
• Input-dependent threshold modulation.

32
Significant Flows

• If number of flows exceeds capacity, then some of
  the flows timeout

• 0 flows in timeout ? Ef 1
• Some flows in timeout ? Ef (0.8 1)
• Most of the flows in timeout. ? Ef a 1/N

33
Algorithm Summary

• Diffusion Early Marking decides whether to mark a
  packet or not as

Where
Remember
M2, b12/3, b21/3
34
Diffusion Mechanisms for Active Queue Management

• Introduction
• Diffusion Early Marking
• Model Optimizations
• Parameter Estimation
• Performance
• Conclusions and Future Work

v
v
v
35
Number of Flows

• The number of significant flows

36
Diffusion Mechanisms for Active Queue Management

• Introduction
• Diffusion Early Marking
• Model Optimizations
• Parameter Estimation
• Performance
• Conclusions and Future Work

v
v
v
v
37
Stability of the Queue

• 100 long lived connections (TCP/Reno, FTP)
• Desired queue size 30 packets

38
Dropping Packets
39
Window Size
40
Changing the number of flows

• 20 new flows every 20 seconds

41
Diffusion Mechanisms for Active Queue Management

• Introduction
• Diffusion Early Marking
• Model Optimizations
• Parameter Estimation
• Performance
• Conclusions and Future Work

v
v
v
v
v
42
Conclusions and Future Work

• Queue length stabilized and controlled without
  adjusting parameters.
• Diffusion mechanism improves the behavior of the
  proposed AQM scheme.
• Future Work
• Optimize the estimation of parameters
• Analyze more traffic scenarios
• Compare with other AQMs
• Use diffusion mechanism in other AQMs

43
Diffusion Mechanisms for Active Queue Management

• Introduction
• Diffusion Early Marking
• Model Optimizations
• Parameter Estimation
• Performance
• Conclusions and Future Work

v
v
v
v
v
v