21
02.12.2004 1 Congestion Control for High Bandwidth-Delay Product Networks D. Katabi (MIT), M. Handley (UCL), C. Rohrs (MIT) – SIGCO MM’02 Presented by Cheng Huang

02.12.20041 Congestion Control for High Bandwidth-Delay Product Networks D. Katabi (MIT), M. Handley (UCL), C. Rohrs (MIT) – SIGCOMM’02 Presented by Cheng

Embed Size (px)

Citation preview

Page 1: 02.12.20041 Congestion Control for High Bandwidth-Delay Product Networks D. Katabi (MIT), M. Handley (UCL), C. Rohrs (MIT) – SIGCOMM’02 Presented by Cheng

02.12.2004 1

Congestion Control for High Bandwidth-Delay Product Networks

D. Katabi (MIT), M. Handley (UCL), C. Rohrs (MIT) – SIGCOMM’02

Presented by Cheng Huang

Page 2: 02.12.20041 Congestion Control for High Bandwidth-Delay Product Networks D. Katabi (MIT), M. Handley (UCL), C. Rohrs (MIT) – SIGCOMM’02 Presented by Cheng

02.12.2004 2

Basics of TCP Congestion Control Bandwidth-delay product

Capacity of the “pipe” between a TCP sender and a TCP receiver

Congestion window (cwnd) sender’s estimation of the capacity

Additive Increase and Multiplicative Decrease (AIMD) algorithm no loss: cwnd = cwnd + s loss: cwnd = cwnd – cwnd/2

Page 3: 02.12.20041 Congestion Control for High Bandwidth-Delay Product Networks D. Katabi (MIT), M. Handley (UCL), C. Rohrs (MIT) – SIGCOMM’02 Presented by Cheng

02.12.2004 3

Motivations

Inadequacy of TCP, as bandwidth-delay product increases Prone to instability

regardless of AQM schemes Inefficient

Fairness concern TCP tends to bias against long RTT flows

Satellite links, wireless links, etc.

Page 4: 02.12.20041 Congestion Control for High Bandwidth-Delay Product Networks D. Katabi (MIT), M. Handley (UCL), C. Rohrs (MIT) – SIGCOMM’02 Presented by Cheng

02.12.2004 4

Design Rationale

NOT an end-to-end approach Using precise congestion signaling Decoupling efficiency and fairness control

Page 5: 02.12.20041 Congestion Control for High Bandwidth-Delay Product Networks D. Katabi (MIT), M. Handley (UCL), C. Rohrs (MIT) – SIGCOMM’02 Presented by Cheng

02.12.2004 5

Features of XCP (eXplicit Control Protocol) Maintains high utilization, small queues, and

almost no drops, as bandwidth/delay increases drop: less than one in a million packets

Maintains good performance in dynamic environment (with many short web-like flows)

No bias against long RTT flows

Page 6: 02.12.20041 Congestion Control for High Bandwidth-Delay Product Networks D. Katabi (MIT), M. Handley (UCL), C. Rohrs (MIT) – SIGCOMM’02 Presented by Cheng

02.12.2004 6

XCP – Sender/Receiver’s Role

Sender Fill the congestion header Update cwnd = max(cwnd + H_feedback, s)

Receiver Copy H_feedback to ACK

Page 7: 02.12.20041 Congestion Control for High Bandwidth-Delay Product Networks D. Katabi (MIT), M. Handley (UCL), C. Rohrs (MIT) – SIGCOMM’02 Presented by Cheng

02.12.2004 7

XCP – Router’s Role

Control Interval Estimation Average RTT

Efficiency Control Maximize link utilization

Fairness Control Achieve fairness among individual flows

Page 8: 02.12.20041 Congestion Control for High Bandwidth-Delay Product Networks D. Katabi (MIT), M. Handley (UCL), C. Rohrs (MIT) – SIGCOMM’02 Presented by Cheng

02.12.2004 8

Control Interval Estimation

Estimation requirement Core stateless Average over flows (not over packets) e.g. two flows have RTTs of 80 ms and 40 ms and

the same cwnd = 10 packets, then average RTT over packets is:

RTTavg = (80*10+40*20)/(10+20) = 53.33 (ms)

Instead, average RTT over flows is:RTTavg = (80*80*10+40*40*20)/(80*10+40*20) = 60 (ms)

Page 9: 02.12.20041 Congestion Control for High Bandwidth-Delay Product Networks D. Katabi (MIT), M. Handley (UCL), C. Rohrs (MIT) – SIGCOMM’02 Presented by Cheng

02.12.2004 9

Control Interval Estimation (2)

Weight of each packet wi = H_rtti * (si / H_cwndi)

Average RTT sum(wi * H_rtti) / sum(wi)

Average cwnd sum(wi * H_cwndi) / sum(wi)

Page 10: 02.12.20041 Congestion Control for High Bandwidth-Delay Product Networks D. Katabi (MIT), M. Handley (UCL), C. Rohrs (MIT) – SIGCOMM’02 Presented by Cheng

02.12.2004 10

Efficiency Controller (EC)

Aggragate feedback (total H_feedback)

alpha, beta: constant value d: control interval (average RTT) S: spare bandwidth Q: persistent queue size

Stability requirement determinesalpha = 0.4; beta = 0.226

Independent of delay, capacity and number of flows

Page 11: 02.12.20041 Congestion Control for High Bandwidth-Delay Product Networks D. Katabi (MIT), M. Handley (UCL), C. Rohrs (MIT) – SIGCOMM’02 Presented by Cheng

02.12.2004 11

Fairness Controller (FC)

Achieve fairness via AIMD algorithm phi > 0, equal throughput increment of all flows phi < 0, throughput decrement proportional to its c

urrent throughput Positive feedback

(1/wi) * (pi/H_rtti) = C1 (constant value) sum(pi/H_rtti) = phi/d

Negative feedback (1/wi) * (ni/H_rtti) = C2 * H_cwndi/H_rtti

sum(ni/H_rtti) = phi/d

Page 12: 02.12.20041 Congestion Control for High Bandwidth-Delay Product Networks D. Katabi (MIT), M. Handley (UCL), C. Rohrs (MIT) – SIGCOMM’02 Presented by Cheng

02.12.2004 12

Fairness Controller (FC) (2)

Bandwidth shuffling h = max(0, gamma*y - |phi|)

gamma = 0.1 y: input traffic

Page 13: 02.12.20041 Congestion Control for High Bandwidth-Delay Product Networks D. Katabi (MIT), M. Handley (UCL), C. Rohrs (MIT) – SIGCOMM’02 Presented by Cheng

02.12.2004 13

Performance Evaluation

Simulation topology I

Page 14: 02.12.20041 Congestion Control for High Bandwidth-Delay Product Networks D. Katabi (MIT), M. Handley (UCL), C. Rohrs (MIT) – SIGCOMM’02 Presented by Cheng

02.12.2004 14

Page 15: 02.12.20041 Congestion Control for High Bandwidth-Delay Product Networks D. Katabi (MIT), M. Handley (UCL), C. Rohrs (MIT) – SIGCOMM’02 Presented by Cheng

02.12.2004 15

Page 16: 02.12.20041 Congestion Control for High Bandwidth-Delay Product Networks D. Katabi (MIT), M. Handley (UCL), C. Rohrs (MIT) – SIGCOMM’02 Presented by Cheng

02.12.2004 16

Performance Evaluation (2)

Simulation topology II

Page 17: 02.12.20041 Congestion Control for High Bandwidth-Delay Product Networks D. Katabi (MIT), M. Handley (UCL), C. Rohrs (MIT) – SIGCOMM’02 Presented by Cheng

02.12.2004 17

The dynamics of XCP (I)

Page 18: 02.12.20041 Congestion Control for High Bandwidth-Delay Product Networks D. Katabi (MIT), M. Handley (UCL), C. Rohrs (MIT) – SIGCOMM’02 Presented by Cheng

02.12.2004 18

The Dynamics of XCP (II)

Page 19: 02.12.20041 Congestion Control for High Bandwidth-Delay Product Networks D. Katabi (MIT), M. Handley (UCL), C. Rohrs (MIT) – SIGCOMM’02 Presented by Cheng

02.12.2004 19

Differential Bandwidth Allocation

Replace FC phi > 0: allocate throughput increment according to flows’ prices

Page 20: 02.12.20041 Congestion Control for High Bandwidth-Delay Product Networks D. Katabi (MIT), M. Handley (UCL), C. Rohrs (MIT) – SIGCOMM’02 Presented by Cheng

02.12.2004 20

Gradual Deployment – A TCP-friendly XCP Separate queues to distinguish TCP and XCP

traffics Calculate average cwnd of TCP traffics by

Update weights to make TCP and XCP fair

Page 21: 02.12.20041 Congestion Control for High Bandwidth-Delay Product Networks D. Katabi (MIT), M. Handley (UCL), C. Rohrs (MIT) – SIGCOMM’02 Presented by Cheng

02.12.2004 21

Conclusion

XCP provides a theoretically sound, yet effective approach to congestion control. It remains excellent performance, independent of link capacity, delay and number of flows.