improving tcp ip performance over wireless network

8/8/2019 Improving Tcp Ip Performance Over Wireless Network

1/25

Improving TCP/IP Performance overImproving TCP/IP Performance overThird Generation Wireless Networks:Third Generation Wireless Networks:M C Chan andM C Chan and RamchandranRamchandran RamjeeRamjee

Bell Labs, Lucent TechnologiesBell Labs, Lucent Technologies

Guided by Shivanth M P

Latha . V Roll No : 60


2/25

AgendaAgenda

y Introduction

y

Windows Regulatory Performance of long-lived TCP flows

y Short-flow differentiation

y

Conclusion


3/25

IntroductionIntroduction2G wireless 2.5G wireless 3G wireless

Phone calls

Voice mailSimple emails

Phone /fax

Voice mailLarge emails

Web browse

Navigation

New updates

Phone/fax

Global roamingLarge emails

High speed web

Video-conferencing

Speed: 10 kb/sec Speed 64-144 Kb/sec Speed:144 kb-2mb/sec

Download 3 min mp3

31-41 min

Download 3 min mp3

6-9 min

Download 3 min mp3

11 sec-1.5 min


4/25

Introduction cont.Introduction cont.

y TCP is most widely used

y Windows regulator that Maximizes

throughput for all buffer sizes (important

metric for long flows)

y Scheduling algorithm to reduce transfer

latency for short flow

y Exploiting user diversity is important


5/25

Simplified 3G networkSimplified 3G network


6/25

TCPTCP WINDOWWINDOW--INGING


7/25

Ack RegulatorAck Regulator


8/25

Windows RegulatorWindows Regulator

Wi = Ni+Yi+1 ------------------- (1)

Yi+1 < Wi ---- ----------------------(2)

Yi+B >= Wi ------------------------ (3)


9/25

Windows RegulatorWindows Regulator--static (WRS)static (WRS)

y On Enqueue of each Ack

Set W=B

Send Ack to sourceWRSAlgorithm

Wi = B and Yi >0 fulfills equation 3

If B


10/25

Windows Regulator Dynamic(WRD)Windows Regulator Dynamic(WRD)

On Dequeue of each data packetY=Y+1

On Enqueue of each Ack packet

Y=Y-1set W=Y+B

Transmit Ack to source

WRD Algorithm

W = Y + B ------ fulfills Equation 3

If B=0, it underflows & violates Equation 2

As W(wrd) >=W(wrs), throughput W/RTT is better for wrd

than wrs


11/25

Windows Regulator withWindows Regulator withAAck Buffer(WRB)ck Buffer(WRB)

On dequeue of each data packet

Y=Y+1

If there is an ack stored in ack buffer then

W=Y+B+Ba

Transmit ack to source

On Enqueue of each ack set

Y=Y+1

W=Y+B+Ba

if W B then no

acks are sent to TCP source.

Ba is ack buffer in reverse direction


12/25

Throughput vs. buffer size(single user)Throughput vs. buffer size(single user)

For TCP selective

ack:

AR performs better

than WRS and DT

WRD is close to

max with Buffer

>15

WRB outperformsall others


13/25

Throughput vs. Buffers Multiple usersThroughput vs. Buffers Multiple users

Similar to Single user

Except AR outperforms

WRD

Absolute performance is

better with multiple users.

8 users are similar to 4

users except gap between

AR and WRD widens.

WRB still performs best.

WRD is best considering

RTT tradeoff.


14/25

Throughput vs. Wired LatencyThroughput vs. Wired Latency

For D < 70ms, WRS is better

than DT. It is degrading>70ms

For D400ms, WRD/WRB are worse

than DT

Increasing RLP buffer sizewill help AR,WRD,WRB


15/25

ThroughputThroughput VsVs LossLossRandom Loss

AR,WRB,WRD perform well for

small amount but start degrading

after 10-3

After 10-2 all algorithms have same

low perf as random error is

dominant factor

Congestion loss

AR,WRB perform better than DT

below congestion loss rate 10-3

WRD/WRS performs poorly thanDT

AR/WRB don't degrade as they

have ack buffer to provide fast

feedback


16/25

Comparison SummaryComparison Summary

y DT and WRS cannot adapt to the large rate anddelay variation in wireless channel

y AR adapts well to the large variations but doesnot perform well with latency as estimation

errors cause throughput degradey WRD performs well against latency but poorly

against congestion loss

y WRB is best and robust against latency andpacket loss


17/25

SchedulingScheduling

y Intra-user scheduling

SFP : short flow priority algorithm

Short flow/Long flow based on data transfer

Reclassification short->long->short

y Inter-user scheduling

PF: Priority Fair queuing

PF-SP: priority fair with Strict PriorityPF-RP: priority fair with Rate priority


18/25

PF schedulerPF scheduler

y Each user reports measured channelconditions to the PF scheduler (RNC)

y User with best channel is selected to

transmit in different time slot, better thanround robin but unfair

y PF weight the current rate achievable byaverage rate received by user and selectuser i with maximum Ri/Ai ,where Ri isthe rate achievable by user andAi isaverage rate .


19/25

PFPF--SP SchedulerSP SchedulerLet set of short flows be SLet set of long flows be L

If S is non empty

select user I with largest Ri

Else ifL is non Empty

select user with largest Ri/Ai

Update Ai for all usersPF-SP SchedulerAlgorithm

y Short flows have higher priority

y It is unfair to long flow.

y Average rate for each user is

maintain over all short and long flows


20/25

PFPF--RP schedulerRP scheduler

Better balance between minimizingShort term latency and fairness

It relatively sacrifices fairness to users

with long flows

Latency reduction improves with more

no of users

Run PF and user i is selected

Run Pf-SP and user j is selected

If Ri>Rj

select user i

else

select user j

Update Ai for all usersThe PF-RP algorithm


21/25

Performance comparisonPerformance comparison

2 users with SNR +4dB and 2

users with -4dB

PF/SFP has better latency

under 15 KB file transferthan PF/FIFO

PF-SP and PF-RP can exploit

differences in channel

conditions


22/25

Performance comparison cont.Performance comparison cont.

All 4 users with same -4dB SNR

PF-SP performs worst for all file

sizes

PF and PF-RP is better than PF-

FIFO

PF/SFP is always better than

PF/FIFO

PF-SP is better or worse than PFdepending on channel condition

PF-RP is the most robust


23/25

ConclusionConclusion

y WRS a common algorithm used in wired

routers perform poorly

y WRB which adapts wireless channel conditions

and performs ack regulation improves 100%over DT

y PF-RP/SFP provides robust performance over

different user channel conditions. 54% over

PF/FIFO


24/25

ReferencesReferences

1 M C Chan and Ramchandran RamjeeImproving TCP/IP performance over Thirdgeneration Wireless networks

Infocom , IEEE 20082 M C Chan and Ramchandran

Ramjee:TCP/IP performance over 3Gwireless link with rate and delay variation

ACM mobicom ,IEEE,2002

3 Wikipedia.org


25/25

Q &A

??

improving tcp ip performance over wireless network

Documents