Channel Allocation for GPRS

From: IEEE Tran. Veh. Technol., Vol. 50, no. 2, 2001.

Author: P. Lin and Y.-B. Lin

CSIE, NTU & CSIE, NCTU

Outline Introduction GPRS architecture Packet data logical channel Channel allocation schemes System model Analysis results Simulation method Performance Conclusion

Introduction

GPRS shares GSM frequency bands with telephone and circuit-switched data traffic

GPRS uses the same TDMA/ FDMA structure as that of GSM to form physical channels

Allocation of channel for GPRS is flexible where one to eight channels can be allocated to a user or one channel can be shared by several users

GPRS architecture

Packet data logical channel Packet data traffic channel (PDTCH)

Employed for transferring of user data Packet broadcast control channel (PBCCH)

Broadcast control Packet common control channel (PCCCH)

The packet random access channel (PRACH) The packet paging channel (PPCH) The packet access grant channel (PAGCH) The packet notification channel (PNCH)

Packet dedicated control channels The packet associated control channel (PACCH) The packet timing advance control channel (PTCCH)

Packet data logical channel (cont’d) Group Name Direction Function

PTC PDTCH Downlink and uplink Data

PBCCH PBCCH Downlink Broadcast

PCCCH PRACH Uplink Random access

PPCH Downlink Paging

PAGCH Downlink Access grant

PNCH Downlink Multicast

PDCH PTCCH Downlink and uplink Timing advance

PACCH Downlink and uplink Associated control

Dynamic allocation: uplink data transfer

Dynamic allocation: downlink data transfer

Channel allocation schemes:•Fixed Resource Allocation (FRA): For a data request of K channels, the BS assigns exact K channels to GPRS packet request •Dynamic Resource Allocation (DRA): For a data request of K channels, DRA allocates at most K channels to the request •Fixed Resource Allocation with Queue Capability (FRAQ) FRAQ_N: a queue for the new calls only FRAQ_H: a queue for the handoff calls only FRAQ_NH: a queue for both new and handoff calls•Dynamic Resource Allocation with Queue Capability (DRAQ) DRAQ_N: similar to FRAQ_N

DRAQ_H: similar to FRAQ_H

DRAQ_NH: similar to FRAQ_NH

A GPRS data request specifies K channels for transmission

The GSM voice call arrival and GPRS packet requests to a cell form Poisson streams with rates and , respectively

The voice call holding time and packet transmission time are exponentially distributed with mean times and , respectively

System model

v p

p/1v/1

: the residence time of voice user at a cell j, which are independent and identically distributed random variables with a general function with mean

jmt ,

vct ,

vcvtuvvcvc eutf ,)( ,,

: the voice call holding time, which is assumed to be exponentially distributed with the density function

jmt ,

η1)( , jmm tf

The timimg diagram

vhλ

bvP

bpP

: the new call blocking probability for the GSM

: dropping/ blocking probability for the GPRS

: voice handoff call arrival rate to a cell

vη : GSM voice user mobility rate

: probability that a GSM voice call is not completed (either blocked or forced to terminate)

ncvP

: the GSM voice call traffic load v

p : the GPRS packet call traffic load

Analytic model for FRA

)]()1(1[

)](1)[1(*

vmfvv

vv*mbvv

vh ufPu

ufP

)]()1(1[

)](1)[()([1)(

*vmfvv

v*mv

*mbvfvbvv

v

vv ufPu

ufufPP+Pη

u

λ=ρ

jmst

t jmmm dtetfsf jm

jm,,

* ,

0,

)()(

where

(1)

(2)

Analytic model for FRA (cont’d)

)!

)(!

()( 1

p

np

v

nv

nnGnp

pv

(3)

(4)

K

CnandCnCKnnnnS pvpvpvFRA 0 ,0,),(

fvvmfvv

vv*mbvv

bv

fvv

hvbvnc

PufPu

ufPP

PPPv

)]()1(1[

)](1)[1(

*

,

state space : FRAS

stationary probability:

Analytic model for FRA (cont’d)

}0,0,)|,{(

)(

K

CnCnCKnnnnn

bv

pvpvpv

npP

}0 0 {(

)(

K

CnC,nC,KnnK)|C,nnn

bp

pvpvpv

npP

])!

)(!

[(

FRA

pv

Sn p

np

v

nv

nnG

ppp where

(5)

(7)

(6)

The iterative algorithm for FRA Step 1: Select an initial value for Step 2: Step 3: Compute and using (2) and (4)-(7) Step 4: Compute using (1) Step 5: If then go to step 2. Otherwise

,go to step 6. Note that is a predefined threshold

say Step 6: The values for , and converge. Compute

from (3)

vhλ

vholdvh λλ ←,

bpPbvP

ncvP

vhλ

vholdvhvh δλλλ ≥- ,

710

vhλ bvPbpP

Analytic model for DRA

}0 ,2

0

,3

≤≤0,≤≤0

,≤23≤0|),,,{(

ClandC

k

CjCi

ClkjilkjiSDRA

vvv ηuM +=

vhvv λλ +=Λ

vvv ηuM +=where

The state transition for DRA

Let’s consider the case when K=3

Analytic model for DRA (cont’d) The balance equations for the Markov process are expressed:

1,,,

,1,,,,1,,,,1

1,,,1,1,,2,,1,3,,,1

,,,321

)1(

)1(2)1(3)1(

)23

lkjip

lkjiplkjiplkjiv

lkjiplkjiplkjiplkjiv

lkjippppppvv

l

kjMi

lkjiM

Analytic model for DRA (cont’d)

Cl

andC

kC

jCi

ClkjilkjiE

0

,2

0,3

≤≤0,≤≤0

,23≤0|),,,(1

1),,,(

,,,Elkji

lkjifvbvbp PPP

: the set of the states where no free channel is available1E

Analytic model for FRAQ

K

CyQCx

QCKyyxS NFRAQ

0 ,0

,x0|),{(_

CKy x0:I Case

Analytic model for FRAQ (cont’d)

CKy x:II Case

QCKy x0:III Case

Analytic model for FRAQ (cont’d)

KC

y

KyQC

KyCxyxfvP

0,

K

CyQCxQCKyxKCyxE 0 ,0 ,|),(2

2),(

,Eyx

yxbpP

KC

y

KyQC

KyCx vv

yxvbv KyCxMKyC

KyCxP

0

,

)1()(

)1(

: a packet request is dropped if the number of free channels is smaller than K

2E

Simulation method We consider a 6x6 wrapped mesh cell

structure The model follows the discrete event

simulation approach

6X6 wrapped mesh cell structure

Performance of FRA ( )

Performance of GPRS data rate: increase as K increase

Effects of packet size :in Fig. 6(b) Effect of voice call arrival: in Fig. 6(c) Effect of voice user mobility: in Fig. 6(d)

voice user mobility has no apparent effect on

bpP

bpP

bpP

Performance of FRA ( )

Effect of packet size: in Fig. 7(b) Effect of voice call arrival: in Fig. 7(c)

packet request have less chance to served as K

increases, and decreases as K increases Effect of voice user mobility: in Fig. 7(d)

high mobility, handoffs are more likely to occur in a voice call,thus for high mobility is larger

ncvP

ncvP

ncvP

Comparison for the FRA and DRA algorithms Performance of

DRA algorithms (with or without queueing) always outperform FRA (with or without queueing)

Performance of

the DRAQ_NH outperforms other algorithms ncvP

bpP

Effect of the variations of the distribution for input parameters

The average number of channels assigned to packet transmission

The average waiting time for the accepted voice call request

Conclusion

The dynamic allocation effectively increases the GPRS packet acceptance rate and queueing mechanisms significantly reduce the voice call incompletion probability