Enhanced power Enhanced power efficient sleep efficient sleep

mode operation mode operation for IEEE 802.16e for IEEE 802.16e based WiMAX based WiMAX Shengqing Zhu, and Tianlei WangShengqing Zhu, and Tianlei Wang

IEEE Mobile WiMAX Symposium, 2007 IEEE Mobile WiMAX Symposium, 2007

Outline Outline • Introduction • System description• An analytical model• Dynamic tuning of initial sleep

window• Performance evaluation• Conclusion

IntroductionIntroduction• IEEE 802.16e (mobile WiMAX) is targeting

for Mobile Subscriber Stations (MSSs)• To efficiently manage energy in IEEE 802.

16e systems– Sleep-mode operation

MotivationMotivation• The analytical results show that there is

a tradeoff between the power consumption and the mean delay– The key of the tradeoff is the initial sleep wi

ndow• Present a heuristic algorithm to tune

the initial sleep window dynamically according to the traffic load

Three type of sleep mode Three type of sleep mode operation in IEEE 802.16eoperation in IEEE 802.16e

• Type 1 power saving class is recommended for connections of non-real-time variable rate traffic

• Type 2 is recommended for connections of real-time variable rate traffic

• Type 3 is recommended for multicast connections as well as for management operations

The overview of the IEEE 802.16e The overview of the IEEE 802.16e power management (Type 1 )power management (Type 1 )

2n until reach its TmaxFixed size

When some data to transmit When some data to transmit in IEEE 802.16e sleep modein IEEE 802.16e sleep mode

BS

MSsL

Tk

a SDU want to transmit

MO

B-T

RF

-

IND

L

MO

B-T

RF

-

IND

Tk-1

awake mode

Dat

a Dat

a Dat

a

BS

MSs

Tk

a SDU want to transmit

LM

OB

-TR

F-

IND

Tk-1

awake mode

Dat

a Dat

a Dat

a Reque

st

Response

delay

QoS is mainly measured by mean delay

System descriptionSystem description• Type 1 power saving class• Downlink • BS transmits all packets addressed to

the MS in its buffer as well as packets arriving to BS during the time the MS receives queued packets

queue

packets arriving

t

System descriptionSystem description• Sleep interval

• Packet arrival process – Poisson process

• arrival rateλ – Ignored listening interval

– Service time of a packet • Generally distributed with PDF

Markov chain model for the sleep Markov chain model for the sleep mechanism of IEEE 802.16emechanism of IEEE 802.16e

• The behaviors of an MS working in power saving mode– bi-dimensional random process {s(t), b(t)} – βi represents the transition probability

An analytical modelAn analytical model

i,0 i+1,0βi

M,0

M,s

1- βM

M-1,0βM-1

βM

i,0

i,s

1- βi

0,s 1,s M,s…

0,0

1 1 1

By means ofBy means ofAn analytical modelAn analytical model

Power consumption • MS can be divided into two parts

– Receiving the expected data packets = ρPr

• ρ = λE[v] (data rate * serving time)– Sleep operation

Delay • When a packet arrives at the BS till the momen

t when the packet is successfully received by the related MS

• j packets addressed to the MS buffered in the BS after the ith sleep state

• The probability that there are j packets buffered after sleep state {i, 0}

Ti…

j packets

t

Delay• The mean number of packets served in the i se

rvice state when j packets buffered after {i, 0}

• Average delay over the packets served in ith service state with j initial packets

• Average delay over all packets

t

j packets

Power Consumption vs. Delay

The power consumption versus initial sleep window

The mean delay versus initial sleep window

Dynamic tuning of initial Dynamic tuning of initial sleep windowsleep window

• If the initial sleep window is small a high traffic load leads to high power consumption– A traffic-aware initial sleep window tuning

should be considered

• After serving all packets buffered, the MS revert to an initial sleep window which is dynamically tuned according to the number of packets served

ExampleExample

Ti+2Ti+1Ti

23

10 packets > 2^3

Ti+2Ti+1

Ti

2 packets < 2^3

Ti Ti+2

t

t

Performance evaluationPerformance evaluation• Simulation tool: NS2 • Simulation time: 50 hours • Average value: 10 runs• Compare with traditional standard:

Tb = 40ms and M = 5

Power consumption comparison for various traffic load

Mean delay comparison for various traffic load

ConclusionConclusion• Present a simple Markov chain model to

investigate the performance of IEEE 802.16e sleep mode operation

• The analytical results show that the power consumption decreases with the initial sleep window but the mean delay increases with it

• Propose a heuristic algorithm, which tunes the initial sleep window dynamically according to the traffic load

• Simulation results show that the proposed algorithm can improve the power consumption significantly

Thanks!!Thanks!!