28
Stochastic Fair Traffic Stochastic Fair Traffic Management for Efficient Management for Efficient and Robust IP Networking and Robust IP Networking Jae Chung Airvana Inc. Chelmsford, MA 01824 Mark Claypool, Robert Kinicki WPI Computer Science Department Worcester, MA 01609 26 th IEEE International Performance Computing and Communications Conference (IPCCC) New Orleans, Louisiana, April 11, 2007

Stochastic Fair Traffic Management for Efficient and Robust IP Networking

Embed Size (px)

DESCRIPTION

Stochastic Fair Traffic Management for Efficient and Robust IP Networking. Jae Chung Airvana Inc. Chelmsford, MA 01824 Mark Claypool, Robert Kinicki WPI Computer Science Department Worcester, MA 01609. 26 th IEEE International Performance Computing - PowerPoint PPT Presentation

Citation preview

Page 1: Stochastic Fair Traffic Management for Efficient and Robust IP Networking

Stochastic Fair Traffic Stochastic Fair Traffic Management for Efficient and Management for Efficient and

Robust IP NetworkingRobust IP Networking

Jae ChungAirvana Inc.

Chelmsford, MA 01824

Mark Claypool, Robert KinickiWPI Computer Science Department

Worcester, MA 01609

26th IEEE International Performance Computing and Communications Conference (IPCCC)New Orleans, Louisiana, April 11, 2007

Page 2: Stochastic Fair Traffic Management for Efficient and Robust IP Networking

IPCCC April 11, 20072

OutlineOutline• Introduction

• SFG Design

• Configuration

• Evaluation

• Summary

Page 3: Stochastic Fair Traffic Management for Efficient and Robust IP Networking

IPCCC April 11, 20073

Internet Congestion ControlInternet Congestion Control• Current Architecture

– TCP : Congestion responsive traffic sources using Additive Increase Multiplicative Decrease (AIMD)

– Drop-Tail IP Router : Implicit congestion feedback controller (packet drop congestion signal)

• Improve Congestion Feedback Control for TCP– Active Queue Management (AQM) at IP Router

• Low queuing delay• Explicit congestion notification (ECN)

• Fairness and Network Protection from non-TCP– Class-based Bandwidth Usage Control– Per-flow Bandwidth Usage Control

Page 4: Stochastic Fair Traffic Management for Efficient and Robust IP Networking

IPCCC April 11, 20074

Complex Light-weightScalable

Preferential-basedPacket Dropping

FQ, SFQ

Pseudo Per-flowManagement

Per-flowManagement

FRED CSFQ, RFQ RED-PD, SFB

StatisticalFlow Monitor

StatisticalPacket Filter

CHOKe

Edge-CoreArchitecture

Scheduling-basedApproaches

Per-Flow Bandwidth ControlPer-Flow Bandwidth Control

SFG

Page 5: Stochastic Fair Traffic Management for Efficient and Robust IP Networking

IPCCC April 11, 20075

Stochastic Fairness GuardianStochastic Fairness Guardian

SFGSFGInIn FilteredFiltered

Congestion Control

Network Protection

IP Router Queue

Drop-Tail/ AQM

Drop-Tail/ AQM

OutOut

dropdrop

TCPTCP

UDPUDP

TCPTCP

UDPUDP

UDPUDP

TCPTCP

dropdrop

TCPTCP

UDPUDP

Page 6: Stochastic Fair Traffic Management for Efficient and Robust IP Networking

IPCCC April 11, 20076

OutlineOutline• Introduction

• SFG Design

• Configuration

• Evaluation

• Summary

Page 7: Stochastic Fair Traffic Management for Efficient and Robust IP Networking

IPCCC April 11, 20077

SFG Design OverviewSFG Design Overview

………

p = 0.05

p (flow3) = 0.00

Level1 Level2 LevelL

Bin1

Bin2

BinN-1

BinN

p (flow2) = 0.02

p (flow1) = 0.03

…p = 0.04

p = 0.02 p = 0.00

p = 0.02

p = 0.03

p = 0.00…

LevelL-1

p = 0.06

Page 8: Stochastic Fair Traffic Management for Efficient and Robust IP Networking

IPCCC April 11, 20078

Multi-Level Hash Bins (1/2)Multi-Level Hash Bins (1/2)• Use multiple hash functions (L)• Each function hashes flows into N bins• Each bin is assigned an equal share (1/N) of the

outbound link capacity (C).• Every epoch (ds), update the forced packet drop

probability for each bin (prob[i][j]):

for i = 0 to L − 1 do for j = 0 to N − 1 do

prob[i][j] = (bytes[i][j] − dsC/N) / bytes[i][j];bytes[i][j] = 0; /* update drop p for all bins */

end for end for

Page 9: Stochastic Fair Traffic Management for Efficient and Robust IP Networking

IPCCC April 11, 20079

Multi-Level Hash Bins (2/2)Multi-Level Hash Bins (2/2)• Each packet arrival, compute the per-flow forced

drop probability (p) for the packet, and update bytes received for each hashed bins:

p = 1;for i = 0 to L − 1 do j = hash(i, packet); p = min(p, prob[i][j]); /* min drop p seen so far

*/ bytes[i][j] = bytes[i][j] + sizeof(packet);end for

• Drop the packet with the computed per-flow drop probability (p)

Page 10: Stochastic Fair Traffic Management for Efficient and Robust IP Networking

IPCCC April 11, 200710

OutlineOutline• Introduction

• SFG Design

• Configuration

• Evaluation

• Summary

Page 11: Stochastic Fair Traffic Management for Efficient and Robust IP Networking

IPCCC April 11, 200711

The Number of Hash BinsThe Number of Hash Bins• Configure maximum and minimum Congestion

Notification Probability (CNP) thresholds (mh, ml) to turn On/Off SFG

if CNP mh then Turn On SFG

if CNP ml then Turn Off SFG

• Find the number of bins (N) such that the capacity of each hash bin is equal to TCP-Friendly Rate (TTCP) at CNP = ml

N = C / TTCP (ml, RTTsys) where,

RTTsys : Estimated average system Round Trip Time

Page 12: Stochastic Fair Traffic Management for Efficient and Robust IP Networking

IPCCC April 11, 200712

The Number of Hash LevelsThe Number of Hash Levels

• False Positive Probability (FPP) Analysis– Given the number of hash bins (N), the number

of hash levels (L) and the estimated number of TCP-Unfriendly flows (B)

– The false positive probability (Pfp) that a TCP-Friendly flow shares all the bins with TCP-Unfriendly flows:

• Use FPP to determine the number of levels.

L

i

k

BkB

iB

ft kik

i

i

Ni

NBNLP

001

11

,,

Page 13: Stochastic Fair Traffic Management for Efficient and Robust IP Networking

IPCCC April 11, 200713

SFG Configuration ExampleSFG Configuration Example• Link Bandwidth = 10 Mbps, RTTsys = 300 ms• CNP Thresholds: mh = 0.02, ml = 0.01• N = 20, B = 1~10, L = ?

Page 14: Stochastic Fair Traffic Management for Efficient and Robust IP Networking

IPCCC April 11, 200714

Unlucky TCP-Friendly Flows?Unlucky TCP-Friendly Flows?• Problem : When hashed, an unlucky TCP-

Friendly flow can always share all the bins with TCP-Unfriendly flows.

• Solution : Use different hashing seed (increment by one) in the next measurement epoch.

• Note : This solution also relaxes the low False Positive Probability (FPP) requirement for long-lived flows (i.e. large file transfer).

Page 15: Stochastic Fair Traffic Management for Efficient and Robust IP Networking

IPCCC April 11, 200715

Measurement Epoch LengthMeasurement Epoch Length• The epoch length should be

– Large enough to avoid control error due to insufficient control data acquisition

– Larger than the effective congestion feedback control system response time to minimize congestion control interference.

• We recommend two seconds for SFG epoch– Approximately twice the upper-bound average

RTT seen on the Internet (1 second) [Choi, INFOCOM 2004]

– The large epoch length, hence slow response time, is acceptable considering the long flow lifetimes of potentially misbehaving flows.

Page 16: Stochastic Fair Traffic Management for Efficient and Robust IP Networking

IPCCC April 11, 200716

OutlineOutline• Introduction

• SFG Design

• Configuration

• Evaluation

• Summary

Page 17: Stochastic Fair Traffic Management for Efficient and Robust IP Networking

IPCCC April 11, 200717

Evaluation OverviewEvaluation Overview• Evaluation Subjects

– Drop-Tail Queue (Baseline)– PI Controller (Hollot+, INFOCOM 2001)– RED-PD (Mahajan+, ICNP 2001),– SFB (Feng+, INFOCOM 2001)– CHOKe (Mitra+, INFOCOM 2000)– SFG, SFG-PI

• Evaluation Objectives– TCP performance– Protection performance– Queuing delay and jitter– Web performance

Page 18: Stochastic Fair Traffic Management for Efficient and Robust IP Networking

IPCCC April 11, 200718

Network Topology and ScenarioNetwork Topology and Scenario

• C = 10 Mbps• Q = 500 Kbytes• RTLD = [60, 1000] ms• Nweb = 300 (Loadoffered = 0.25)

– Web session setting (H-Campos+, MASCOTS 2003)

Sizeavg= 5KB, Shape = 1.2, Tavg_think = 7sec (expo distribution)

• Nftp_bw = 50• Nftp_fw = 10 50 100 200 400 (every 200 sec)

400 200 100 50 10 (every 200 sec)• Ncbr = 5

– 2 Mbps CBR (1.2 Mbps VBR) from 100 to 1700 sec

• Simulation time = 2000 sec

r1 r2

s

s

s

s

d

d

d

d

Q = 500 pkts

C = 10 Mbps

Page 19: Stochastic Fair Traffic Management for Efficient and Robust IP Networking

IPCCC April 11, 200719

Queue Configurations (1/2)Queue Configurations (1/2)

• RED-PD– RED : qmin = 50, qmax = 300, pmax = 0.15, wq =

0.002

– PD : RTTtarget = 100 ms, Windowflow_monitor_history = 5,

Tflow_unmonitor = 15 sec, Ratedrop_threshold = 0.005,

pmax_inc_step = 0.05

• CHOKe– RED : Same as above – Packet Filter : Divide RED’s queue threshold range

(qmax - qmin) into 5 even sub-regions and apply 2i+1 drop comparisons for an incoming packet, where i = {0, 1, 2 ,3, 4} is the sub-region ID.

Page 20: Stochastic Fair Traffic Management for Efficient and Robust IP Networking

IPCCC April 11, 200720

Queue Configurations (2/2)Queue Configurations (2/2)

• SFB– BLUE (inside each SFB bin) :

pinc_step = 0.005, pdec_step = 0.001, Tfreeze = 100 ms

– Flow Monitor :

L = 3, N = 20, punresp_detect = 0.98,

Tpenalty_box = 15 ms, Thash_switch = 20 sec

• SFG-PI– PI : KP = 0.71× 10−5, KI = 2.8116 × 10−5

– SFG : L = 3, N = 20, mh = 0.02, ml = 0.01, ds = 2 sec

Page 21: Stochastic Fair Traffic Management for Efficient and Robust IP Networking

IPCCC April 11, 200721

System ThroughputSystem Throughput

Offered Load1.2 1.4 1.7 1.4 1.21.11.0 1.01.1

# of FTP 100 200 400 200 100 50 1010 50 400

1.7

Page 22: Stochastic Fair Traffic Management for Efficient and Robust IP Networking

IPCCC April 11, 200722

Unresponsive CBR ThroughputUnresponsive CBR Throughput

Offered Load1.2 1.4 1.7 1.4 1.21.11.0 1.01.1

# of FTP 100 200 400 200 100 50 1010 50 400

1.7

Page 23: Stochastic Fair Traffic Management for Efficient and Robust IP Networking

IPCCC April 11, 200723

Queuing Delay and JitterQueuing Delay and JitterOffered Load 1.2 1.4 1.7 1.4 1.21.11.0 1.01.1

# of FTP 100 200 400 200 100 50 1010 50 400

1.7

Page 24: Stochastic Fair Traffic Management for Efficient and Robust IP Networking

IPCCC April 11, 200724

Web Object Service TimeWeb Object Service Time

Offered Load1.2 1.4 1.7 1.4 1.21.11.0 1.01.1

# of FTP 100 200 400 200 100 50 1010 50 400

1.7

Page 25: Stochastic Fair Traffic Management for Efficient and Robust IP Networking

IPCCC April 11, 200725

OutlineOutline• Introduction

• SFG Design

• Configuration

• Evaluation

• Summary

Page 26: Stochastic Fair Traffic Management for Efficient and Robust IP Networking

IPCCC April 11, 200726

SummarySummary• IP Router Queue Management Taxonomy

• Stochastic Fairness Guardian (SFG)– A lightweight Statistical Packet Filter– Flexible deployment with Drop-Tail or AQM– Practical configuration guidelines– Performs comparable to or better than complex

flow monitoring mechanisms (RED-PD, SFB).

Page 27: Stochastic Fair Traffic Management for Efficient and Robust IP Networking

IPCCC April 11, 200727

Additional ContributionAdditional Contribution

• Confirms [Le+, SIGCOMM 2003] result that ECN degrades Web service time at a high offered load (1.2). This is because the congestion notification probability (CNP) is significantly higher than that of packet drop congestion notification system, causing more TCP SYN packet drops.

Page 28: Stochastic Fair Traffic Management for Efficient and Robust IP Networking

Stochastic Fair Traffic Stochastic Fair Traffic Management for Efficient and Management for Efficient and

Robust IP NetworkingRobust IP Networking

Jae ChungAirvana Inc.

Chelmsford, MA 01824

Mark Claypool, Robert KinickiWPI Computer Science Department

Worcester, MA 01609

26th IEEE International Performance Computing and Communications Conference (IPCCC)New Orleans, Louisiana, April 11, 2007