“A dynamic rate allocation technique for wireless communication systems”

March 15th 2004

1Department of Electronics and Telecommunications

“A dynamic rate allocation technique

for wireless communication systems”

Romano Fantacci Full Professor

Francesco Chiti Ph.D.

Daniele Tarchi Ph.D.

Department of Electronics and TelecommunicationsUniversity of Florence

Via di S. Marta, 3I-50139 Florence, ITALY

E-mail: {fantacci,chiti,tarchi}@lenst.det.unifi.it

March 15th 2004


Outline1. Motivations

– 3G systems features

– UMTS-HSDPA

2. ALC Protocol Proposal– Physical channel – Proposed rate allocation protocol– System block diagram– Traffic sources – Overall Markov model– Analytical model

3. Numerical Results– N(), T() theoretical and simulated– Gain

4. Conclusions and further developments

March 15th 2004


Future wireless networks features1. Motivations

• Global coverage by means of: – Efficient internetworking with existing wireless and wired

standards by resorting to a cooperative approach rather than competitive

– High mobility and variable traffic load management through dynamic Radio Resource Management policies (RRM):

• cell planning• system reconfiguration

• Services integration– Multimedia traffics with real time (voice, audio/video streaming)

and data (Web services, Data Base queries) applications – Different Quality of Service (QoS) requirements (bandwidth,

error rate, delays)– Asymmetric connections handling

March 15th 2004


1. Motivations

Universal Mobile Telecommunications System

144 Kbps500 Km/h

384 Kbps120 Km/h

2 Mbps10 Km/h

UMTS/HSDPAcdma2000/1xEV

DOIEEE802.11x

TETRA2

March 15th 2004


1. Motivations

High Speed Downlink Packet Access3GPP Release 5 (2001) arranges a further downlink access scheme to handle asymmetric, high bit rate, bursty data services in an indoor environment.

This purpose could be achieved by, eventually, joint selection of the following strategies:

1. Adaptive Modulation and Coding (AMC) schemes2. Hybrid Automatic Repeat reQuest (H-ARQ) techniques3. Fast scheduling algorithms4. Multiple Inputs Multiple Outputs (MIMO) channel modelling5. Fast Cell Selection (FCS) algorithms

Constraints:• Peak bit rate up to 10 Mbs • No QoS degradation

March 15th 2004


Upper Bound of MC Scheme

1,00E-07

1,00E-06

1,00E-05

1,00E-04

1,00E-03

0 5 10 15

SNR [dB]

BE

R

QPSK+Turbo Code

16 QAM+Turbo Code

2. ALC Protocol Proposal

Modulation and Coding SchemesModulation and Coding Schemes

I II III

M Rc R [bit/symb]

I

II QPSK 1/2 1

III 16QAM 1/2 2

March 15th 2004



Physical Channel ModelPhysical Channel Model

Channel ouput

0

1

23

4

5

6

7

89

10

11

12

0 5 10 15 20 25 30 35 40 45 50

Time [ms]

rec

eiv

ed

Po

we

r [m

W]

GOOD GOOD

BAD BAD

RP

TP

March 15th 2004


Tgood , Tbad : exponentially distributed

PT : threshold power level

PR : average received power level


Exponential (memoryless) hypothesis [Gupta84]

R

T

P

P

2

1

D

goodf

T

2

1

D

badf

eT

c

fvfD

0


March 15th 2004



Discrete Memoryless Channel (DMC)

Bad0

Good 1

01r

10r

00r 11r

channel transition probability between i state and j state within a slot:

• exponentially (geometrically) distributed• mean value related to received signal power and user

mobility

jir ,


March 15th 2004


1. Base Station (BS) manages downlink streams according to a FIFO scheduling policy

2. Whenever an End User (EU) is selected, BS discretely monitors EU physical channel conditions

3. Depending on channel state, a proper AMC scheme is chosen

4. BS allocates to this EU both:• Dedicated Physical Channel (DPCH)• Downlink Shared Channel (DSCH) with a variable shared

capacity


Proposed ProtocolProposed Protocol

March 15th 2004



System Block Diagram (uplink/downlink)

Core NetworkBase Station

Buffer Scheduler

Wireless Channel

Channel Monitor

Mobile User

Proposed ProtocolProposed Protocol

March 15th 2004


Traffic Sources

• Poisson packet arrivals

• Poisson message arrivals:

1. Modified Geometric distribution of packet within each message

2. Pareto message length (constant length packets): simplified Web traffic

3. Pareto packet length and Exponential packet inter-arrivals: real Web traffic)

3. approaches 2. in the presence of an high capacity CN connection


System ModelSystem Model

March 15th 2004


Traffic Sources

• Poisson packet arrivals:

• Batch message arrivals Geometrical message length:

Pareto message length:

ka k packets arrival probability within a

slot

k

IFFT

ez

zM

ezaezA jj

1

ek

ak

k !

pz

zpzM

1

1

CDF Pareto

01

k

k

kkk

xF

zxFxFzM



March 15th 2004


0,0r00a0 0,1 0,2

r00a0r00a0

0,k-2 0,k-1 0,k

r00a0r00

a0

0,k+1

r00a1 r00a1 r00a1 r00

a1

r00a2 r00a2 r00

a2 r00a2 r00

a2

r00a1r00a1 r00

a1r00a1

1,000a0 1,1 1,2

r11a1r11a1

1,k-2 1,k-1 1,k

r11a1r11a1

1,k+1

r11a2 r11a2 r11

a2 r11a2

r11a3 r11a3 r11a3 r11a3 r11a3

r11a2r a r a2r a

r

r00a0

r11a1

r11a0 r11

a0

11


System ModelSystem ModelDT Embedded Markov chain model [Neuts89]

Vectorial state (i,j): •i : status of the transmission channel •j : number of packets in the queue

March 15th 2004



Steady State Equations

0,01,00

1,00,11,12,10

1,1,1

0,11,12,10

0,10,01,00

0,0,0

paparpapaparp

papaparpaparp

kik

k

iikkik

k

iik

kkik

k

iikik

k

iik

probability of being in i phase with j queued packetsjip ,

1

0 0,

i jjijpN

0

,110

,00j

jj

j pRpR 0,10,0 ppN

T

probability of having k packets arrivalska


March 15th 2004


)1())(1()()()()()(

)1())(1()()()()()(2

0,11,10,01,01,11,101,01,12

1

20,10,11,10,10,00,010,10,0

20

zprprprzzzGzPzGzrzGrzzP

zprprprzzzGzPzGrzzGrzzP

arrival generating function

k

kk zazG

0

)(

)1()1()1( 10 PPPN average queued packets

0,10,02pp

TTINT

average queuing time (by Little

formula)


Transformed Domain Equations


March 15th 2004


• 3GPP standard compliant: IPv6 fast backbone:

- maximum message length equal to 5 MB (truncated Pareto pdf)

- packet length equal to 1.5 KB

Time slot (TTI) equal to 2 ms

Bit rate equal to 1.92 Mbps

• Worst case multipath fading: and r01 = r10 = 0.2 (duty cycle = 0.5)

• Infinite shared memory buffer length: no dropping effect

• ARQ policy belonging to GB class (RTT<TTI)

• Poutage equal to 5%

3. Numerical Results

Operative AssumptionsOperative Assumptions

March 15th 2004


NePSi: a Network Protocol Simulator• NePSi (Network Protocol Simulator) is a Discrete Event Simulator • It is based on C++ programming language• Object oriented programming is used in order to model different

entities in the system

• S. Nannicini, T. Pecorella, L. S. Ronga, “IneSiS: Integrated Network Protocols and Signal Processing Simulator”, Sixth Baiona Workshop 1999, Vigo, Spain.

• Available at http://lenst.det.unifi.it/INeSiS/ under GNU License.

3. STF 179 Proposal

March 15th 2004



Poisson packet arrival: N

HSDPA Gain: improving transport bit rate or network capacity (QoS) or decreasing on board device complexity

MD1 vs HSDPA

0

2

4

6

8

10

12

0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4

Arrival Rate [packet per slot]

Av

era

ge

Pk

t n

um

be

r

MD1 Th

HSDPA Th

HSDPA Sim

March 15th 2004


MD1 vs HSDPA

0,003

0,006

0,009

0,012

0,015

0,018

0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4


Qu

eue

wai

tin

g t

ime

[s]

MD1 Th

HSDPA Th

HSDPA Sim

HSDPA Gain: lowering expected delay (QoS)


Poisson packet arrival: T

March 15th 2004


Impact of Channel slotted monitoring

0

2

4

6

8

10

12

0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4


Av

era

ge

Pk

t n

um

be

r

Continuos monitoringDiscrete monitoring

Moderate impact on protocol efficiency


Poisson packet arrival:

March 15th 2004


MD1 vs HSDPA

0

40

80

120

160

200

240

280

320

360

0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4


Av

era

ge

Pk

t n

um

be

r

MD1 Th

HSDPA Th

HSDPA Sim


Geometrical Batch message arrival: N

March 15th 2004


MD1 vs HSDPA

0

5000

10000

15000

20000

25000

30000

0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4


Av

era

ge

Pk

t n

um

be

r

MD1 SimHSDPA Sim


Pareto Batch message arrival (3GPP): N

March 15th 2004



Increasing HSDPA Gain along with traffic burstiness

GTraffic models comparison:

Gain Comparison

0

10

20

30

40

50

60

70

80

90

100

0,0 0,2 0,4 0,6 0,8 1,0


Ga

in [

%]

Gain Poisson

Gain Geometric

Gain Pareto

March 15th 2004



pdfNTraffic models comparison:

N and ak are statically similar: few queued messages (protocol efficiency)

March 15th 2004


Unreliabell vs Reliable Services

0

2

4

6

8

10

0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4


Ave

rag

e P

kt n

um

ber

MD1 no ARQ

MD1 ARQ

HSDPA no ARQ

HSDPA ARQ

ARQ protocols less affect HSDPA performance


Poisson packet arrival: N

March 15th 2004


Conclusion• High QoS applications (high bit rate, time sensitiveness) feasibility

within 3G networks has been investigated

• Following 3GPP recommendations, as to novel HSDPA scheme, a new protocol has been proposed

• Based on physical channel state observation, a dynamic bandwidth is allocated to users

• Protocol efficiency has been tested under several traffic models, including Web services models (LRD)

• A remarkable gain has been highlighted, if compared with M/D/1 systems

4. Conclusion & developments

March 15th 2004


More accurate channel monitoring (3 MCS allocation)

Further Developments4. Conclusion & developments

• State 0: 4-QAM, Rc=1/2



1 2 3 r33r11

r22

r21

r12

r32

r23

March 15th 2004


Enhanced Link Adaptation Algorithm

Steady State Equations (3 states)

0,11,12,10

2,10,21,22,23,20

2,2,2

0,21,22,22,00

1,2

0,01,00

1,00,11,12,10

1,1,1

0,11,12,10

0,10,01,00

0,0,0

ppaparpppaparp

pppapar

paparppaparp

ppaparpaparp

kik

k

iikik

k

iik

kik

k

ii

kik

k

iikik

k

iik

kik

k

iikik

k

iik


March 15th 2004


Publications[1] F. Chiti, L. Caponi, R. Fantacci: “Dynamic Bandwidth Allocation in Wireless Communications Systems”, in Proc. of AIRO 2002.

[2] F. Chiti, L. Caponi, R. Fantacci: “An Efficient Rate Allocation Technique based on Channel Status Observation for Wireless Communication Systems”, in Proc. of IEEE WCNC 2004.

[3] F. Chiti, L. Caponi, R. Fantacci: “A Dynamic Rate Allocation Technique for Wireless Communication Systems ”, in Proc. of IEEE ICC 2004.

[4] F. Chiti, L. Caponi, R. Fantacci: “A Dynamic Radio Resources Allocation Technique for Wireless Communication Systems ”, submitted to Trans. on Vehic. Tech.

Founded Research ProjectsETSI STF 179 on TETRA Release 2 TEDS Adaptive Link Control


“A dynamic rate allocation technique for wireless communication systems”

Documents