parameterizing pi congestion controllers

16
Parameterizing PI Congestion Controllers Ahmad T. Al-Hammouri, Vincenzo Liberatore, Michael S. Branicky Case Western Reserve University Stephen M. Phillips Arizona State University April 3, 2006 Support by: NSF CCR-0329910, Department of Commerce TOP 39-60- 04003, NASA NNC04AA12A, and an OhioICE Training grant

Upload: zev

Post on 06-Jan-2016

29 views

Category:

Documents


0 download

DESCRIPTION

Parameterizing PI Congestion Controllers. Ahmad T. Al-Hammouri, Vincenzo Liberatore, Michael S. Branicky Case Western Reserve University Stephen M. Phillips Arizona State University April 3, 2006 - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Parameterizing PI Congestion Controllers

Parameterizing PI Congestion Controllers

Ahmad T. Al-Hammouri, Vincenzo Liberatore, Michael S. Branicky

Case Western Reserve University

Stephen M. PhillipsArizona State University

April 3, 2006

Support by: NSF CCR-0329910, Department of Commerce TOP 39-60-04003, NASA NNC04AA12A, and an OhioICE Training grant

Page 2: Parameterizing PI Congestion Controllers

Ahmad Al-Hammouri Parameterizing PI Congestion Controllers FeBID’06

2/16

Contributions of the Paper

Complete stability region for PI Presents examples that show

Different stable PI parameters exhibit widely different control performance

Neglecting delays in control design leads to unstable systems

Page 3: Parameterizing PI Congestion Controllers

Ahmad Al-Hammouri Parameterizing PI Congestion Controllers FeBID’06

3/16

TCP-AQM Control Loop

x(t)=cwnd/RTT

tf

q(t)f(q(t))q`=Σx(t) - C

p(t)

tb

Controller G(s)

On Ackcwnd += 1/cwndOn losscwnd /= 2

( )( )sdB

es s

Plant

P(s)=[Hollot et al 2001]

Page 4: Parameterizing PI Congestion Controllers

Ahmad Al-Hammouri Parameterizing PI Congestion Controllers FeBID’06

4/16

PI AQM [Hollot et al 2001]

PI vs RED Control signal:

.

Frequency transfer fn:

.

0( ) ( ) ( )

tp iu t k e t k e d

( )i

pk

G s ks

q(s)G(s) P(s)

q0

+ _e u

P

PI

Integral term eliminatesthe steady-state error

Page 5: Parameterizing PI Congestion Controllers

Ahmad Al-Hammouri Parameterizing PI Congestion Controllers FeBID’06

5/16

Parameterizing PI AQM

Problem: Determine the entire region of stabilizing

kp and ki values Objectives:

Stable closed-loop system Enhanced closed-loop performance

Steady-state error, convergence time, overshoot

Challenges: Delays

Page 6: Parameterizing PI Congestion Controllers

Ahmad Al-Hammouri Parameterizing PI Congestion Controllers FeBID’06

6/16

Contributions of the Paper

Complete stability region for PI Presents examples that show

Different stable PI parameters exhibit widely different control performance

Neglecting delays in control design leads to unstable systems

Page 7: Parameterizing PI Congestion Controllers

Ahmad Al-Hammouri Parameterizing PI Congestion Controllers FeBID’06

7/16

p i{(k , k ) :

( )lim 1} =

( )( )

p i

s

k s k B

s s s

Complete Stability Region SR [Silva et al 2005]

SR = (S0 \ SN) \ SL

S0 : Stability region for

the delay-free system P0(s) =

SN : .

p i

0

{(k , k ) :

lim|G(s)P (s)| 1}s

( )( )

B

s s S0 \ SNS0

Page 8: Parameterizing PI Congestion Controllers

Ahmad Al-Hammouri Parameterizing PI Congestion Controllers FeBID’06

8/16

Determination of SL

Set of kp and ki’s that destabilizes the closed-loop for delays less than d0

For given kp and ki

Find d that gives the blue curve If (d ≤ d0), (kp,ki ) SL

Else, (kp,ki ) SL

Sweep (kp,ki ) S0

-1

Im{s}

x

d++

Re{s}

Nyquist Plot ofG(s)P(s)

Page 9: Parameterizing PI Congestion Controllers

Ahmad Al-Hammouri Parameterizing PI Congestion Controllers FeBID’06

9/16

Complete Stability Region SR

SR = (S0 \ SN) \ SL

S0 \ SN

SR

Page 10: Parameterizing PI Congestion Controllers

Ahmad Al-Hammouri Parameterizing PI Congestion Controllers FeBID’06

10/16

Contributions of the Paper

Complete stability region for PI Presents examples that show

Different stable PI parameters exhibit widely different control performance

Neglecting delays in control design leads to unstable systems

Page 11: Parameterizing PI Congestion Controllers

Ahmad Al-Hammouri Parameterizing PI Congestion Controllers FeBID’06

11/16

Example 1

N = 75; d0 = 0.15 sec; C = 1250 pkt/sec

*[Hollot et al 2001]

*

Page 12: Parameterizing PI Congestion Controllers

Ahmad Al-Hammouri Parameterizing PI Congestion Controllers FeBID’06

12/16

Complete stability region for PI Presents examples that show

Different stable PI parameters exhibit widely different control performance

Neglecting delays in control design leads to unstable systems

Contributions of the Paper

Page 13: Parameterizing PI Congestion Controllers

Ahmad Al-Hammouri Parameterizing PI Congestion Controllers FeBID’06

13/16

PIP Controller [Heying et al 2002]

q(s)G(s) P(s)

q0

+ _

e u

Kh

+ _

PI

PIP Kh

Page 14: Parameterizing PI Congestion Controllers

Ahmad Al-Hammouri Parameterizing PI Congestion Controllers FeBID’06

14/16

Example 2 [Heying et al 2002]

N = 60; d0 = 0.22 sec; C = 1250 pkt/sec

*[Heying et al 2002]

Page 15: Parameterizing PI Congestion Controllers

Ahmad Al-Hammouri Parameterizing PI Congestion Controllers FeBID’06

15/16

Future Work

Conduct packet-level simulations (ns-2) Define a “Networks Performance”

objective function Optimize the objective function over the

stability region Analyze the queue nonlinearity (i.e.

truncation)

Page 16: Parameterizing PI Congestion Controllers

Ahmad Al-Hammouri Parameterizing PI Congestion Controllers FeBID’06

16/16

Thank You

Questions Comments