channel allocation for the gprs design and performance study huei-wen ferng, ph.d. assistant...
Post on 20-Dec-2015
218 views
TRANSCRIPT
Channel Allocation for the GPRSDesign and Performance Study
Huei-Wen Ferng, Ph.D.
Assistant ProfessorDepartment of Computer Science and Information Engineering (CSIE)Nation Taiwan University of Science and Technology (NTUST)Wireless Communications and Networking Engineering (WCANE) LabURL: mail.ntust.edu.tw/~hwferngE-mail: [email protected]
2
NTUST
WCANE LabCSIE
Outline
Introduction Channel allocation schemes System model and assumptions Performance study and numerical examples Conclusions
3
NTUST
WCANE LabCSIE
GPRS Architecture
4
NTUST
WCANE LabCSIE
Service Requirements
Blocking vs. forced termination From the viewpoint of users, one may feel more
uncomfortable when an on-going call is abruptly terminated than directly getting blocked before his service.
Generally speaking, less forced termination than blocking.
Delay-sensitive vs. non-delay-sensitive Voice is more sensitive than data.
5
NTUST
WCANE LabCSIE
Design Principles
Channel reservation It privileges handoff calls.
Priority Priority among buffers
Differentiation between voice and data requests Service priority between new voice calls and handoff voice calls
Differentiation between new calls and handoff calls Buffering strategy
Allows more net input rates Threshold control
Throttles different rates of new calls and handoff calls
6
NTUST
WCANE LabCSIE
Channel Allocation Schemes
An example of dynamic allocation of the uplink data transfer
7
NTUST
WCANE LabCSIE
Basic Dynamic Channel Allocation (DCA): For a data request, DCA allocates at most n channels to the request. For a voice call, only one channel is allocated.
Five DCAs are proposed based on service priority, threshold control, channel reservation, and buffering strategies.
Channel Allocation Schemes
: no priority for voice and data buffers, no threshold control.
: higher priority for voice buffer, no threshold control.
: higher priority for voice buffer with threshold control.
: similar to the 2n
d
scheme with handoff the highest pri.
: similar to the 2n
d
scheme with channel reservation.
NCQPDCA
NCNQPDCA
TCQPDCA
NCQHPDCA
NCQPGDCA
8
NTUST
WCANE LabCSIE
CAS1 Scheme
1
C
1dB
vB 1
Voice buffer
Data buffer
New voice (handoff) calls
New data packets
FIFO
This scheme is proposed for reference.
9
NTUST
WCANE LabCSIE
1
C
1dB
vB 1
Voice buffer
Data buffer
New voice (handoff) calls
New data packets
High priority
Low priority
When buffer is empty.
CAS2 Scheme
10
NTUST
WCANE LabCSIE
New voice (handoff) calls
New data packets
CAS3 Scheme
When buffer is emptyHandoff call first.Then, new voice call.
1
C
1dB
vB 1
Voice buffer
Data buffer
High priority
Low priority
11
NTUST
WCANE LabCSIE
1
C
1dB
vB 1
Voice buffer
Data buffer
High priority
Low priority
New data packets
New voice (handoff) calls
CAS4 Scheme
vT
When buffer is empty
Handoff voice calls
Blocked when exceeding the threshold
New voice calls
12
NTUST
WCANE LabCSIE
New voice (handoff) calls
New data packets
New call and handoff call
Handoff call
When buffer is empty.
CAS5 Scheme
1
C
GCC
1dB
vB 1
Voice buffer
Data buffer
High priority
Low priority
13
NTUST
WCANE LabCSIE
GSM (new) voice call and GPRS (new) data packet arrive according to Poisson processes with ratesλv and λd, respectively.
GSM (new and handoff) voice call holding time, GPRS packet transmission time, and GSM user dwelling time follow exponential distributions with mean 1/μv , 1/μd and 1/η, respectively..
Static data users are assumed for simplicity.
Each GSM user moves to any adjacent cell in a uniform manner; same traffic load as well as same number of channels are assumed to any cell. (resulting in homogenous cells).
System Model and Assumptions
14
NTUST
WCANE LabCSIE
Simulation Environment
Square cell structure 6x6 wrapped mesh cells Homogeneous cells
15
NTUST
WCANE LabCSIE
Define state space Write balance equations based on the state
transition diagram Use the recursive approach to obtain results
Analysis of Channel Allocation Scheme
16
NTUST
WCANE LabCSIE
ddvhvni
n
iidhnn
CAS
BbBbBbi
Cy
CxCiyxbbbyyx
S
0 ,0 ,0 ,0
0 ,0|,,,,,,0
1
3
Define state space
17
NTUST
WCANE LabCSIE
State transition diagram
for otherwise 0,
if 1,
11for otherwise 0,
if 1,
1
1
njnCiyx
n-jjCiyx
n
ii
pi
n
ii
pi
ji
jiji if ,0
if ,1
For convenience, let us define two sets of indicator functions
18
NTUST
WCANE LabCSIE
State transition diagram
dhnknnkkkk
dhninniiii
dhninniiii
dhnndhnn
dhnn
bslbsbyyylx
n
k
k
l
l
sdk
bbbyyyx
n
idi
bbbyyyx
n
idpi
bbbyyxbbbyyx
bbbyyx
n
i
n
idpidi
Pyk
Pyi
P
PxP
Piyx
,,,,,,,,2
1
1 0
,,,,,,,,1
,,,,,,,,11
,,,,,,1,,,,,,1
,,,,,1 1
11
11
11
11
1
)1(
)1(
)1(
)(
0,0 when : 1 1
dhnd
n
ii bbbCiyxCase
19
NTUST
WCANE LabCSIE
State transition diagram
20
NTUST
WCANE LabCSIE
Blocking probabilities
fttftfftfft
vbtvbfvbfvb
PPPP
PPPP
)1(
)1(
)]([),,,,,(,,,,,,
3,1
1
vhvCASdhnn
dhvnBbbSbbbyyx
bbbyyxftfvbf PPP
)1(3,1
1),,,,,(
,,,,,,
ftfh
Sbbbyyxbbbyyxh
ftt P
Pb
P CASdhnn
dhvn
)1(3,1
1),,,,,(
,,,,,,
vbfv
Sbbbyyxbbbyyxn
vbt P
Pb
P CASdhnn
dhvn
)]([),,,,,(
,,,,,,
3,1
1
ddCASdhnn
dhvnBbSbbbyyx
bbbyyxdd PP
21
NTUST
WCANE LabCSIE
Delay times
)1(
)1(
)1(
31
,1
31
,1
31
,1
),,,,,,(,,,,,
),,,,,,(,,,,,
),,,,,,(,,,,,
ddd
Sbbbyyxbbbyyxd
d
ftfh
Sbbbyyxbbbyyxh
hv
vbfv
Sbbbyyxbbbyyxn
nv
P
Pb
W
P
Pb
W
P
Pb
W
CASdhnn
dhnn
CASdhnn
dhnn
CASdhnn
dhnn
Using Little’s formula:
22
NTUST
WCANE LabCSIE
Number of channels (C ) 7
Voice buffer size 7
Maximum available number of channels for data packets (n)
3
Voice holding time
Data transmission time
Data arrival rate
Parameters Setting
v100/1
v25
1/1 v
23
NTUST
WCANE LabCSIE
Performance Measures
Blocking probability for a new voice call Forced termination probability for a handoff voice
call Data packet dropping probability Delays Cost comparisons among different schemes
24
NTUST
WCANE LabCSIE
The Effect of Data Buffering (on new voice blocking probability)
25
NTUST
WCANE LabCSIE
The Effect of Data Buffering (on forced termination probability)
26
NTUST
WCANE LabCSIE
The Effect of Data Buffering (on data dropping probability)
27
NTUST
WCANE LabCSIE
The Effect of Data Buffering (on delays of data packet)
Delays of new voice calls and handoff voice calls are similar to probabilities.
28
NTUST
WCANE LabCSIE
The Effect of Data Buffering
Data buffer size affects little to new call blocking probability and handoff call forced termination probability, except CAS1.
Increasing data buffer size greatly improves data dropping probability.
The effects on delays of new voice calls and handoff calls are similar to blocking probability and forced termination probability, respectively. Increasing data buffer size raises data packet delay time.
29
NTUST
WCANE LabCSIE
The Effect of Threshold Control(on blocking/termination probability)
Decreasing Tv makes new voice call blocking increase and improves forced termination probability.
30
NTUST
WCANE LabCSIE
The Effect of Threshold Control(on data dropping probability)
Because lower value of Tv permits fewer queued new voice calls in the system;therefore, data packets have a better chance to be served, thus data dropping probability decreases.
31
NTUST
WCANE LabCSIE
The Effect of Threshold Control(on delays of new voice calls and handoff voice calls)
32
NTUST
WCANE LabCSIE
The Effect of Threshold Control(on delays of data packets)
33
NTUST
WCANE LabCSIE
Schemes Comparison (new voice blocking probability for various traffic load)
The best to worst schemes are CAS2, CAS3, CAS4, and CAS5.
34
NTUST
WCANE LabCSIE
Schemes Comparison (forced termination probability for various traffic load)
The best to worst schemes are CAS5, CAS4, CAS3 and, CAS2.
35
NTUST
WCANE LabCSIE
Schemes Comparison (data dropping probability for various traffic load)
CAS4 performs best in terms of data dropping probability.
36
NTUST
WCANE LabCSIE
Schemes Comparison (delays of new voice calls for various traffic load)
37
NTUST
WCANE LabCSIE
Schemes Comparison (delays of handoff voice calls for various traffic load)
38
NTUST
WCANE LabCSIE
Schemes Comparison (delays of data packets for various traffic load)
39
NTUST
WCANE LabCSIE
Effect of data user mobility and computation illustration on channel reservation
The performance of voice calls is not sensitive to variation of ηd because voicecalls have higher precedence over data packets.
When CG increases, performance of new voice calls and data packets becomes poor.
40
NTUST
WCANE LabCSIE
Effect of different data traffic models
The exponential distribution may not be appropriate in modeling data traffic.
Instead, the Pareto distribution can be used to capture the nature of data traffic.
tt
tf ,)(1
41
NTUST
WCANE LabCSIE
Cost comparisons(fixedτ3)
ddvbftf PPPc 321 Cost function:
42
NTUST
WCANE LabCSIE
Cost comparisons(fixedτ2)
43
NTUST
WCANE LabCSIE
Cost comparisons (fixedτ1)
44
NTUST
WCANE LabCSIE
Conclusions
We conclude:Buffering for both voice calls and data packets reduces blocking probability, forced termination probability, and data dropping probability but it increases delay times. The threshold control is an effective approach to reduce forced termination and data dropping probabilities. But it enlarges new voice call blocking probability.
We have examined and compared the improvement of channel allocation schemes using four techniques.
45
NTUST
WCANE LabCSIE
Conclusions
Finally, we suggest scheme CAS3 and CAS4 to be used in the GPRS system.
Although reservation greatly improves forced termination probability and delay time for handoff voice call, it causes the performance of new voice call and data service poor.
46
NTUST
WCANE LabCSIE
The EndThank You!