Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer...

32
Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer Science Columbia University in the City of New York

Transcript of Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer...

Page 1: Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer Science Columbia University in the City of New York.

Active Queue Management: Theory, Experiment and

Implementation

Vishal MisraDept. of Computer ScienceColumbia University in the

City of New York

Page 2: Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer Science Columbia University in the City of New York.

Collaborators

•C.V. Hollot, Don Towsley: UMass Amherst

•Victor Firoiu: Nortel Networks

•Kevin Jeffay, Nguyen-Long Le, Don Smith: UNC Chapel Hill

Page 3: Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer Science Columbia University in the City of New York.

Outline

• Investigating rate based control• Implementation of PI controller

– Hardware– Software

• Experiment– Performance evaluation under

generated web traffic

Page 4: Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer Science Columbia University in the City of New York.

TCPdynamic

queuedynamic

AQM

RN

p

W q

time delayR secs

MGT Fluid-Flow Model

“oscillatory behavior increases with increasing round-trip time”

Page 5: Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer Science Columbia University in the City of New York.

TCPdynamic

AQM

RN

p

W

time delayR secs

Kelly

“oscillatory behavior decreases with increasing round-trip time”

x

Page 6: Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer Science Columbia University in the City of New York.

Paradox?

• MGT model : control based on queue length (q)

• Kelly model : control based on arrival rate (x)

Rate Feedback p = g(x) =B

Cx

Page 7: Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer Science Columbia University in the City of New York.

Utilization with different B

Page 8: Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer Science Columbia University in the City of New York.

)()(

))((2

)(1)(

2

tWRN

tx

RtxgRtW

RtW

)('2

)(2

)( 02

0

0

RtxC

xWg

RW

tWRW

tW refRN

Rate Feedback p = g(x) =

RN

p

W

time delayR secs

x

0

2

1

RWs -

p(t - R)

linearization

B

ref

C

xx

'2

20 gR

W

Page 9: Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer Science Columbia University in the City of New York.

L(s)

p

-

p(t - R)

0

1

02

20

0

2

20

22

2

'2)(

0

RWs

C

xWB

CRNW

RWs

gRNW

sL

B

refRN

gB

Crefxx

sRe

B

refRN

C

xWW

02

02

where W0 satisfies:

rate feedback loop

(*)

Page 10: Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer Science Columbia University in the City of New York.

Cxref

unstable for > 0.3

Stability (B=1)

Stability distance of Nyquist plot from –1+j0

N=60 flowsC=3750 packets/sec

Page 11: Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer Science Columbia University in the City of New York.

Simulations at RTT =300 ms

unstable for > 0.3

N=60 flowsC=3750 packets/sec

Page 12: Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer Science Columbia University in the City of New York.

Parabolic rate feedback B = 2

12

22

22

22

)(0

02

30

0

02

20

2

0

1

02

20

sRW

xWRC

NW

RWs

C

xW

CRNW

RWs

C

xW

CB

RNW

sLrefR

NrefR

N

B

B

refRN

Where, W0 satisfies:

022 220

230

40

2 CWxWxW refRN

refRN

Page 13: Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer Science Columbia University in the City of New York.

Multiple Equilibria (Throughput)

Page 14: Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer Science Columbia University in the City of New York.

Multiple Equilibria (Stability)

Page 15: Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer Science Columbia University in the City of New York.

Cxref Stability (B=2)

unstable for > 0.8

N=60 flowsC=3750 packets/sec

Page 16: Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer Science Columbia University in the City of New York.

Simulations at RTT = 300 ms

N=60 flowsC=3750 packets/sec

Page 17: Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer Science Columbia University in the City of New York.

Implementation and Experiments

Page 18: Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer Science Columbia University in the City of New York.

Implementing PI controller

PI

q(t) p(t)

qref

Integral controller, regulates router buffer to someoperator controlled value qref

Page 19: Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer Science Columbia University in the City of New York.

Hardware Implementation

• Active collaboration with two vendors on implementing PI on a router– Nortel Networks: Next generation

edge router– Cisco: IOS on the 3260 platform

Page 20: Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer Science Columbia University in the City of New York.

Transitioning from theory to practice (Nortel)

• Theory, Simulations: Worry about computations at one output queue, for a single class of traffic

• Practice: Typical router has M (~ 512) queues, E (~ 8) classes

Page 21: Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer Science Columbia University in the City of New York.

Speed Issues

Consider a 10 GBps router, 1000 byte average packet size

•Theory: Sampling interval (say) 1 ms: computational overhead spread over 40000 packets: “lightweight computations”

•Practice: Sampling interval 1ms, MxE (512x8) computations: spread over 10 packets: significant overhead!

Page 22: Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer Science Columbia University in the City of New York.

Memory issues

• Theory: One drop/marking probability needs to be maintained

• Practice: MxE values have to be maintained!

• Hardware designers unwilling to allot memory real estate for AQM (relatively small part of a router)

Solution: Discretize [0,1] and use small precomputed tables

Page 23: Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer Science Columbia University in the City of New York.

Architecture

0101

Append pointer to probability lookup table

0101 .12767

Table for class i

packet

Packet from priority class i

Lookup probability

Droppingmodule

Small (~8) number of tables used with finite (~ 16) entries

Page 24: Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer Science Columbia University in the City of New York.

Open research issues

• How do you discretize [0,1] ?– Linear is clearly not the answer:

operating region typical below 0.2

• Given a typical operating range of p : what performance metric do we optimize? What is the cost function?

Page 25: Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer Science Columbia University in the City of New York.

Software Implementation of PI

• “Tuning RED for Web Traffic”, Sigcomm 2000– Implemented RED on a software router (the

ALTQ system running on FREEBSD)– Compared performance of RED and FIFO

(Droptail) on a testbed with generated Web traffic: studied request completion latency

– Conclusions: RED normally does not help, difficult to tune for scenarios when it can help

(read: “RED only possibly helps in really extreme cases and even here it's hard as hell to get the settings right”)

AQM bad idea?

Study of AQM at UNC

Page 26: Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer Science Columbia University in the City of New York.

Handwaving explanation

pRTTK

B FIFO

pRTT

KBRED

More losses, more retransmissions, more timeouts..-> higher latency!

Page 27: Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer Science Columbia University in the City of New York.

UNC Testbed

Page 28: Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer Science Columbia University in the City of New York.

PI Implementation on ALTQ

• PI added as a module to ALTQ at UNC• Issues: no floating point arithmetic allowed,

need to be careful about saturation, integer overflows!

• Sigcomm 2000 experiments repeated under (nearly) identical conditions with PI as third mechanism

• PI tuned using formula given in Infocom 2000 paper

Page 29: Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer Science Columbia University in the City of New York.

Plot of CDF of response time of requests (80% load)

Cu

mu

lati

ve p

rob

ab

ility

Response time (ms)

Page 30: Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer Science Columbia University in the City of New York.

Plot of CDF of response time of requests (100% load)

Cu

mu

lati

ve p

rob

ab

ility

Response time (ms)

PI, qref=20 FIFO, RED

PI, qref=200

Page 31: Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer Science Columbia University in the City of New York.

Plot of CDF of response time of requests (110% load)

Cu

mu

lati

ve p

rob

ab

ility

Response time (ms)

PI, qref=20

PI, qref=200

FIFO, RED

Page 32: Active Queue Management: Theory, Experiment and Implementation Vishal Misra Dept. of Computer Science Columbia University in the City of New York.

Preliminary conclusions

• AQM may not be bad after all: PI/20 performs significantly better for short objects under heavy load

• Experiments run with packet dropping, not ECN

• ECN experiments planned: performance should improve dramatically over FIFO