Upload
alegria-martinez
View
37
Download
0
Embed Size (px)
DESCRIPTION
An Energy Efficient Sleep Scheduling Considering QoS Diversity for IEEE 802.16e Wireless Networks. Jen- Jee Chen, Jia -Ming Liang and Yu- Chee Tseng Department of Computer Science National Chiao Tung University, Hsin -Chu 30010, Taiwan. IEEE ICC 2010. Speaker: Wun -Cheng Li. Outline. - PowerPoint PPT Presentation
Citation preview
An Energy Efficient Sleep Scheduling Considering QoS Diversity for IEEE
802.16e Wireless Networks
Speaker: Wun-Cheng LiIEEE ICC 2010
Jen-Jee Chen, Jia-Ming Liang and Yu-Chee Tseng
Department of Computer ScienceNational Chiao Tung University, Hsin-Chu 30010, Taiwan
2
• Introduction•Tank-Filling (TF) allocation Scheme•Performance evaluation•Conclusion
Outline
3
•Power management is one of the most important issues in IEEE 802.16e wireless networks.
• IEEE 802.16e standard defines three types of power saving classes (PSCs) for flows with different QoS characteristics.
Type IType IIType III
Introduction
4
•Three IEEE 802.16e power saving classes
Type I :
Type II :
Type III :
Introduction
…
TS_init
2 x TS_init TL
(TS_max)
TL TS
…
TSnormal operation
sleep windows
listening windows
5
Introduction
Bandwidth
MS1
MS3
MS2
Frame Data Burst
•Different MSSs with different service requirements and delay constraints
Sleep less
Waste of bandwidth
6
Problem•Each MS stays in sleep periods as much as possible
but not to violate the QoS requirements•How to increase bandwidth utilization and to reduce
energy consumption for MSs?
Frame Data Burst
MS1
MS3
MS2
Bandwidth
7
Goals•Less power consumption•Higher bandwidth utilization
• Each Mi has a data arrival rate of τi bits/frame and each data arrival has a delay bound of Di frames
8
Network Environment
: Denote service packet’s number per frame : Bandwidth requirement of Mi per sleep cycle
i
Mi (τi , Di)
M1(0.2Ω , 4)
Active frame
Sleep frame
γ1 = τ1 x 4 = 0.8 Ω
M1
M1
γ1= 0.8Ω
τ1 = 0.2Ω
9
TF allocation Scheme•This Tank-Filling (TF) algorithm proposed
three steps.Step AStep BStep C
•Step ABS first sort MSSs by their delay bounds. Without loss
of generality, let D1 ≤ D2 ≤ · · ≤ Dn.
Supposing that = is known and ≤ D1, we set, , i = 2..n, as follows:
10
TF allocation Scheme
SiT
: Delay bound of Mi
: Sleep cycle of MiS
i
i
T
D
•Step AThis property helps make MSSs’ sleeping behaviors
regular and increase the overlapping of their listening windows.
11
TF allocation Scheme
M1
M2
M3
D1= 2 frames
D2= 5 frames
D3= 9 frames
overlapping
12
TF allocation Scheme•Step A
D1= 2 frames
D2= 5 frames
D3= 9 frames
ST1 = Tbasic
ST2 = 2 Tbasic
ST3
= 4 Tbasic
= Tbasic = 2 frames = × ( 5 / ) => = 2 Tbasic
= × ( 9 / ) => = 2 = 4 Tbasic
ST1ST2
ST1ST1
ST2
ST3
ST2ST2
ST3ST2
1cycle = 4 Tbasic = 8 frames
M1
M2
M3
13
TF allocation Scheme•Step B
Calculate the bandwidth requirement of Mi per by γi = τi ×
Scheduling and Select the least number of active frames an leave the least remaining
resource in the last frame.
SiT S
iT
1 32
M1
M2
1 32
M3
M4
OiT L
iT
: Listening windows of MiL
O
i
i
TT : offset of Mi
14
TF allocation Scheme•Step B
R[k, l]: the remaining bandwidth in l-th frame of k-th Tbasic
R(4,3)R(1,1) R(4,2)R(4,1)R(3,3)R(3,2)R(3,1)R(2,3)R(2,2)R(2,1)R(1,3)R(1,2)
(1st Tbasic)
(2nd Tbasic)
(3rd Tbasic)
(4th Tbasic)
Tbasic = 3 frames
0.5Ω
1.25 Ω
0.4Ω
0.4Ω
2.5 Ω
M1
M2
M3
M4
M5ST5 = 12 frames = 4Tbasic
ST3 = 6 frames = 2Tbasic
ST2 = 6 frames = 2Tbasic
ST1 = 3 frames = Tbasic
ST4 = 6 frames = 2Tbasic
0.5Ω 0.5Ω0.5Ω0.5Ω
1.25Ω 1.25Ω
0.4Ω 0.4Ω
0.4Ω 0.4Ω
2.5Ω
15
TF allocation Scheme•Step B
0.5Ω
1.25 Ω
0.4Ω
0.4Ω
2.5 Ω
R[1,1]=0.5Ω R[2,1]=0.5Ω R[3,1]=0.5Ω R[4,1]=0.5Ω
For M1, (1 Tbasic)B1,1=B1,4=B1,7=B1,10=0.5Ω
= 1 = 1OT1
LT1
R[k, l]: the remaining bandwidth in l-th frame of k-th Tbasic
(1st Tbasic) (2nd Tbasic) (3rd Tbasic) (4th Tbasic)
M1
M2
M3
M4
M5
16
TF allocation Scheme•Step B
For M2, (2 Tbasic)1. 0.5Ω + 1.25Ω - (0.5Ω + 1.25Ω) = 0.25Ω2. 1.25Ω − 1.25Ω = 0.75Ω
B2,1=B2,7=0.5Ω , B2,2=B2,8=0.75Ω
=1 = 2OT2
LT2
R[2,1]=0.5ΩR[1,2]=0.25Ω R[3,2]=0.25Ω R[4,1]=0.5ΩR[1,1]=0R[3,1]=0
0.5Ω
1.25 Ω
0.4Ω
0.4Ω
2.5 Ω
R[k, l]: the remaining bandwidth in l-th frame of k-th Tbasic
M1
M2
M3
M4
M5
(1st Tbasic) (2nd Tbasic) (3rd Tbasic) (4th Tbasic)
R[1,3]=0.75Ω R[3,3]=0.75Ω
17
TF allocation Scheme•Step B
For M3, (2 Tbasic)1. FULL2. R[1,2]= 0.25Ω < 0.4Ω3. R[1,3]= 1 Ω > 0.4Ω, // 1 Ω -0.4 Ω =0.6 Ω4. R[2,1]= 0.5 Ω > 0.4Ω, // 0.5 Ω -0.4 Ω =0.1 Ω5. R[2,2]= 1 Ω > 0.4Ω, // 1 Ω -0.4 Ω =0.6 Ω6. R[2,3]= 1 Ω > 0.4Ω, // 1 Ω -0.4 Ω =0.6 ΩB3,4=B3,10=0.4Ω = 4 = 1OT3
LT3
R[2,1]=0.1ΩR[1,2]=0.25Ω R[3,2]=0.25Ω R[4,1]=0.1Ω
0.5Ω
1.25 Ω
0.4Ω
0.4Ω
2.5 Ω
R[k, l]: the remaining bandwidth in l-th frame of k-th Tbasic
M1
M2
M3
M4
M5
(1st Tbasic) (2nd Tbasic) (3rd Tbasic) (4th Tbasic)
18
TF allocation Scheme•Step B
For M4, (2 Tbasic)1. FULL2. R[1,2]= 0.25Ω < 0.4Ω3. R[1,3]= 1 Ω > 0.4Ω, // 1 Ω -0.4 Ω =0.6 Ω4. R[2,1]= 0.1 Ω < 0.4Ω5. R[3,3]= 1 Ω > 0.4Ω, // 1 Ω -0.4 Ω =0.6 ΩB4,3=B4,9=0.4Ω = 3 = 1
OT4LT4
0.5Ω
1.25 Ω
0.4Ω
0.4Ω
2.5 Ω
R[k, l]: the remaining bandwidth in l-th frame of k-th Tbasic
R[2,1]=0.1ΩR[1,2]=0.25Ω R[3,2]=0.25Ω R[4,1]=0.1Ω
R[1,3]=0.6Ω R[3,3]=0.6Ω
M1
M2
M3
M4
M5
(1st Tbasic) (2nd Tbasic) (3rd Tbasic) (4th Tbasic)
19
TF allocation Scheme•Step B
For M5, (4 Tbasic)Frames 1~12 are enough for the data requirement of M5, 2.5 ΩB5,2=0.25Ω , B5,3=0.6Ω , B5,4=0.1Ω , B5,5=1Ω , B5,6=0.55Ω = 2 = 5
OT5LT5
R[1,2]=0.25Ω R[3,2]=0.25Ω R[4,1]=0.1Ω
R[3,3]=0.6Ω
R[1,3]=0R[2,1]=0R[2,2]=0
R[2,3]=0.45Ω
0.5Ω
1.25 Ω
0.4Ω
0.4Ω
2.5 Ω
M1
M2
M3
M4
M5 R[k, l]: the remaining bandwidth in l-th frame of k-th Tbasic
0.6Ω 1.65 Ω
(1st Tbasic) (2nd Tbasic) (3rd Tbasic) (4th Tbasic)
R[2,1]=0.1Ω
R[1,3]=0.6Ω
20
TF allocation Scheme•Step B
For M5, (4 Tbasic)Frames 1~12 are enough for the data requirement of M5, 2.5 ΩB5,2=0.25Ω , B5,3=0.6Ω , B5,4=0.1Ω , B5,5=1Ω , B5,6=0.55ΩB5,3=0.6Ω , B5,4=0.1Ω , B5,5=1Ω , B5,6=0.8Ω = 3 = 4
OT5LT5
R[1,2]=0.25Ω R[3,2]=0.25Ω R[4,1]=0.1Ω
R[3,3]=0.6Ω
R[1,3]=0R[2,1]=0R[2,2]=0
R[2,3]=0.2Ω
0.5Ω
1.25 Ω
0.4Ω
0.4Ω
2.5 Ω
M1
M2
M3
M4
M5 R[k, l]: the remaining bandwidth in l-th frame of k-th Tbasic
0.6Ω 1.9 Ω
(1st Tbasic) (2nd Tbasic) (3rd Tbasic) (4th Tbasic)
R[2,1]=0.1Ω
R[1,3]=0.6Ω
•Step CHere we adopt an exhausted search by setting = 1..D1
and trying to find the sum of the total number of active frames of all MSSs over a windows of frames.
Then leading to the least number of active frames is chosen as Tbasic
21
TF allocation Scheme
sT1
*sTn
ST5 = 12 frames = 4Tbasic
ST3 = 6 frames = 2Tbasic
ST2 = 6 frames = 2Tbasic
ST1 = 3 frames = Tbasic
ST4 = 6 frames = 2Tbasic
ST5 = 8 frames = 4Tbasic
ST3 = 4 frames = 2Tbasic
ST2 = 4 frames = 2Tbasic
ST1 = 2 frames = Tbasic
ST4 = 4 frames = 2Tbasic
Tbasic = 2 frames
22
Simulation Parameters
Parameter ValueMSSs Mi 5 to 45
data rate τi 1000 ~ 3000 bits / frame
delay bound Di 10 ~ 200 frames
available bandwidth Ω 80000 bits (16Mbps) / frame
• C++
23
Simulation Parameters•A. Effects of n
24
Simulation Parameters•B. Effects of Maximum Delay Bound
n = 20
25
Simulation Parameters•C. Effects of System Bandwidth
n = 20
Conclusions
26
•This algorithm leads to each MSS can sleep more and the total power consumption of the system is significantly reduced.
•The proposed scheme is easy to implement and compatible to the standard.
27
Thank you!