以流量為基礎之IEEE 802.16e睡眠排程機制
A Load-based Power Saving and
Scheduling Scheme in IEEE 802.16e
國立暨南國際大學
資訊工程系 楊峻權
2010.05.04
Outline

Introduction





Wireless Standards, IEEE 802.16e/m
Power Saving Techniques
IEEE 802.16e/m Power Saving Class
Related Work
Load-based Power Saving

LBPS-Aggr, LBPS-Split, LBPS-Merge

Performance Evaluation

Conclusion
2
Wireless Standards
Wide Area Network (WAN)
802.16e/m
Nomadic
802.20
Mobile
802.21
Handoff
802.22
WRAN
2, 2.5, 3G
Cellular
Metropolitan Area Network (MAN)
802.16/WiMax
Fixed Wireless MAN
Local Area Network (LAN)
802.11
Wi-Fi
Personal Area Network (PAN)
802.15.1
Bluetooth
802.15.3
802.15.4
Zigbee
3
IEEE 802.16 Standards (1)
Standard
802.16
802.16a
802.16-2004
802.16e
應用模式
固定式應用（取代寬頻設備）
移動式應用
應用方向
Last Mile & Backhaul
Mobile Device
頻段
10~66 GHz
2~11 GHz
2~6 GHz
傳輸條件
LOS
NLOS
NLOS
傳輸速率
32~134 Mbps
75 Mbps
15 Mbps
調變技術
QPSK, 16QAM,
64QAM
移動性
固定性
固定性
移動性
傳輸距離
1~3 Miles
4~6 Miles (Max 30 Miles)
1~3 Miles
QPSK, 16QAM, 64QAM QPSK, 16QAM, 64QAM
(採256 Subcarrier OFDM) (採257 Subcarrier OFDM)
4
IEEE 802.16 Standards (2)
5
IEEE 802.16e


Newly developed broadband wireless
communication technology
MSS is battery-powered


An effective power-saving strategy is
necessary for extending the operation
time
Periodically turn off the transceiver to
save power (Sleep Mode)
6
IEEE 802.16e MAC protocol


Frequency division duplex (FDD) mode and
time division duplex (TDD) mode
Downlink: from the BS to MSSs


Point-to-multipoint broadband wireless access
Uplink


Multiple MSSs share one slotted uplink channel
via TDD on a demand basis for voice, data, and
multimedia traffic
The BS handles bandwidth allocation by assigning
uplink slots based on requests from MSSs
7
IEEE 802.16e Service classes

Unsolicited Grant Service (UGS)

Real-Time Polling Service (rtPS)

Non-Real-Time Polling Service (nrtPs)

Best Effort (BE)
IEEE 802.16e  IEEE 802.16m (1Gbps, 4G)
8
IEEE 802.16m Service classes

Real-time constant bit-rate (e.g., VoIP without silence suppression)

Extended real-time variable bit-rate (e.g., VoIP with silence suppression)

Real-time variable bit-rate (e.g., MPEG video)

Non-real time variable bit-rate (e.g., FTP, HTTP)

Best effort (e.g., E-mail)
RT-CBR
ERT-VR
RT-VR
NRT-VR
Periodic
Periodic
Periodic
Dynamic Dynamic
Packet size
Fixed
Fixed/
dynamic
Dynamic
Dynamic Dynamic
Delay sensitivity
High
High
High
Data generated
interval
Middle
BE
Low
9
Power Saving Techniques (1)

Application layer



Transport layer


Load partitioning (computation performed at BS)
Reduce # of transmissions for operations (e.g. via
data compression)
Reduce # of retransmissions
Network layer

Power efficient routing through a multi-hop
network
10
Power Saving Techniques (2)

Data link layer



MAC layer



Reduce # of packet errors at a receiving node
Automatic Repeat Request (ARQ) and Forward
Error Correction (FEC)
Sleep scheduling protocols
Cycle the radio between its on and off power
states
Physical layer

Proper hardware design techniques
11
IEEE 802.16e Power Saving




Three types of Power Saving Class (PSC)
Type I: MSS doubles its next sleep period
if no packets are sent or received
Type II: MSS repeats the sleeping and
listening periods in a round-robin fashion
Type III: MSS sleeps for the predefined
period and then returns to normal
operation
12
IEEE 802.16e Type I PSC
13
IEEE 802.16e Type II PSC
14
IEEE 802.16e Type III PSC
15
IEEE 802.16m Type IV PSC
16
IEEE 802.16m PSC
17
Related Work (1)

Performance analysis (mainly PSC Type I)

For downlink traffic by semi-Markov chain
(IEEE Comm. Mag. 2005)


For downlink & uplink by Poisson traffic
pattern (IEEE Comm. Mag. 2006)
Hyper-Erlang distributed inter-arrival time
(IEEE WCNC 2007)

Optimal selection of PSC I and II (IEEE
WCNC 2007)
18
Related Work (2)

Adaptive power saving mechanisms


Adjusting the waiting time before entering
the sleep mode
Adjusting the initial and final sleep
windows (IEEE Globecom 2006)

Delay-based sleep scheduling

Latest enhancements (IEEE Trans. VT 2009,
2010)
19
Load-based Power Saving

Weakness of PSC I and PSC II


Traffic modeling and Measurement


Poisson arrival process (uplink & downlink)
MSS’s load  sleep cycle length


Exponential increase or constant pattern
Data accumulation threshold (1 time frame)
BS responsible for sleep schedule
20
LBPS in light load
21
LBPS in heavy load
22
LBPS-Aggr Protocol
23
LBPS Mathematics (1)
24
LBPS Mathematics (2)
Data_TH = one time frame of data
Prob_TH = 0.8 in the simulation
25
Problem with LBPS-Aggr



Unrealistic assumption of synchronized
sleep cycle for all MSSs
Low utilization of mini-slots in a time
frame
Two enhancements


LBPS-Split
LBPS-Merge
26
LBPS-Split Protocol (1)
27
LBPS-Split Protocol (2)
28
Features of LBPS-Split




Dividing MSSs to separate groups
All MSSs with the same length (K*) of
the sleep cycle
Is it possible to use different value of
K* for different MSS?  LBPS-Merge
Schedulability for different K*
29
LBPS-Merge Protocol (1)
2
30
LBPS-Merge Protocol (2)
31
Simulation Study
# of MSS (one BS)
10, 20, 40, 80
# of mini-slots in a time frame 160
Value of Prob_TH
0.8
Packet size
1 mini-slot
7
Simulation time
1*10 time frames
Type I initial sleep interval
2 time frames
Type I maximal sleep interval
2 time frames
Length of listening window
1 time frame
0
9
Load distribution among MSS Equal, 8:2, Random
32
Performance Criteria

Power Saving Efficiency (PSE)
S
Sleep_window size
A
Awake_window size
K time frames
...
S
K-1 time frames


A
...
1 time
frame
PSE = (K-1)/K
Access delay
33
Power Saving Efficiency (1)
10 MSSs, Equal Load
LBPS-Split
LBPS-Merge
LBPS-Aggr
Standard Type I
1
Power Saving Efficiency
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
Total load
0.7
0.8
0.9
0.95
34
Power Saving Efficiency (2)
10 MSSs , 8:2 Load
LBPS-Split
LBPS-Merge
LBPS-Aggr
Standard Type I
1
Power Saving Efficiency
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0.95
Total load
35
Access Delay (1)
10 MSSs, Equal Load
LBPS-Split
LBPS-Merge
LBPS-Aggr
Standard Type I
45
Delay (time frame)
40
35
30
25
20
15
10
5
0
0.1
0.2
0.3
0.4
0.5
0.6
Total load
0.7
0.8
0.9
0.95
36
Access Delay (2)
10 MSSs, 8:2 Load
LBPS-Split
LBPS-Merge
LBPS-Aggr
Standard Type I
35
Delay (time frame)
30
25
20
15
10
5
0
0.1
0.2
0.3
0.4
0.5
0.6
Total load
0.7
0.8
0.9
0.95
37
LBPS-Split: Impact of # MSS (1)
LBPS-Split, Equal Load
PSE
10 MSSs
20 MSSs
40 MSSs
80 MSSs
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
Total load
0.7
0.8
0.9
0.95
38
LBPS-Split: Impact of # MSS (2)
LBPS-Split, 8:2 Load
PSE
10 MSSs
20 MSSs
40 MSSs
80 MSSs
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
Total load
0.7
0.8
0.9
0.95
39
LBPS-Merge: Impact of # MSS (1)
LBPS-Merge, Equal Load
PSE
10 MSSs
20 MSSs
40 MSSs
80 0 MSSs
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
Total load
0.7
0.8
0.9
0.95
40
LBPS-Merge: Impact of # MSS (2)
LBPS-Merge, 8:2 Load
PSE
10 MSSs
20 MSSs
40 MSSs
80 MSSs
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
Total load
0.7
0.8
0.9
0.95
41
Impact of load distribution (1)
LBPS-Split, 10MSSs
Equal Load
8:2 Load
Random
1
0.9
0.8
0.7
PSE
0.6
0.5
0.4
0.3
0.2
0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
Total load
0.7
0.8
0.9
0.95
42
Impact of load distribution (2)
LBPS-Merge, 10MSSs
PSE
Equal Load
8:2 Load
Random
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
Total load
0.7
0.8
0.9
0.95
43
Conclusion & Future Work


Load-Based Power Saving

Traffic Modeling & Measurement

LBPS-Aggr, LBPS-Split, LBPS-Merge

Better power saving efficiency
Future work

Integrated real-time and non-real-time

More general traffic modeling
44