1. Department of Information Engineering University of Padova, ITALY Mathematical Analysis of IEEE 802.11 Energy Efficiency. A note on the use of these ppt slides: We’re making these slides freely available to all, hoping they might be of use for researchers and/or students. They’re in PowerPoint form so you can add, modify, and delete slides (including this one) and slide content to suit your needs. In return for use, we only ask the following: If you use these slides (e.g., in a class, presentations, talks and so on) in substantially unaltered form, that you mention their source. If you post any slides in substantially unaltered form on a www site, that you note that they are adapted from (or perhaps identical to) our slides, and put a link to the authors webpage: www.dei.unipd.it/~zanella Thanks and enjoy!

2. Department of Information Engineering University of Padova, ITALY Mathematical Analysis of IEEE 802.11 Energy Efficiency {andrea.zanella, depe}@dei.unipd.it Andrea Zanella, Francesco De Pellegrini WPMC 2004, 12-15 September 2004 Special Interest Group on NEtworking & Telecommunications

3. WPMC'04 Abano Terme, Padova (Italy) 12-15 September 2004 Motivations  Wireless ad-hoc networks are becoming more and more popular  Self-organization  Mobility  Portability  IEEE 802.11 offers native support for ad-hoc networking  Single cell managed by means of Distributed Coordination Function (DCF)‏  Terminals are battery-powered: energy consumption is a primary issue!  Energy consumption in transmission and reception is of the same order of magnitude [Feeney 01]  The carrier-sense mechanism (CSMA/CA) reduces collision probability but draws energy [Stemm 97]  Cost of sensing is exacerbated by transmissions occurring during the backoff  Also collisions and alien traffic involve an energetic cost

4. WPMC'04 Abano Terme, Padova (Italy) 12-15 September 2004 Aim of the study  Goal  Providing a complete statistical description of the energy spent  Characterize the impact of RTS/CTS on energy consumption  Provide a mathematical tool for the design of energy-aware algorithms Case study  Reference scenario [Bianchi2000]  Ad hoc network with n saturated IEEE 802.11 terminals Single-hop network No hidden or exposed node problem Heavy traffic conditions (saturation)‏ All terminals have always a packet ready for transmission

5. WPMC'04 Abano Terme, Padova (Italy) 12-15 September 2004 Energy Model  Linear energetic model  Energy is drawn proportionally to the time spent in each mode [Feeney]  Each operating mode is associated to a different energetic coefficient Transmitting (  )‏ Receiving (  R )‏ Sensing (  S )‏ Virtual Sensing (  0 )‏ RTS CTS DATA ACK T RTS T SIFS T CTS T DATA T ACK T DIFS RTS CTS DATA ACK T SIFS RTS NAV T NAV A)‏ B)‏ C)‏ A B C  Energy spent during SIFS periods is neglected

6. WPMC'04 Abano Terme, Padova (Italy) 12-15 September 2004 Detailing the Energy Consumption Overall energy spent for successful packet delivery Energy spent in non-colliding transmission Energy spent in colliding transmissions Energy spent during backoff Number of collisions before success Energy spent in each collision  Hypothesis  are i.i.d. and independent of E T  Probability of collision p independent of the system state [Bianchi01]

7. WPMC'04 Abano Terme, Padova (Italy) 12-15 September 2004 Detailing E T & E Tc,j  E T : Energy required for transmitting a packet with success Basic Access RTS/CTS  E Tc,j : Energy spent during packet collision Basic Access RTS/CTS T EIFS

8. WPMC'04 Abano Terme, Padova (Italy) 12-15 September 2004 Detailing E B  E B : Energy spent in backoff  W r : total number of tick periods spent in backoff  Tick Period time between two successive decrements (tick) in the backoff countdown process Idle channel: countdown 1 per time slot Busy channel: freeze until the channel returns idle for a DIFS, then resume countdown   j : energy spent in each tick period

9. WPMC'04 Abano Terme, Padova (Italy) 12-15 September 2004 Detailing  j  During a tick period a node can be  sensing the radio channel  receiving a valid packet intended for that node  discarding a valid packet for other destinations  listening collided transmission on the channel Idle Channel Busy Channel sense receive listen discard

10. WPMC'04 Abano Terme, Padova (Italy) 12-15 September 2004 Putting al pieces together...  Moment generating function for the energy spent by each node in the network

11. WPMC'04 Abano Terme, Padova (Italy) 12-15 September 2004 Case Study  Lucent WaveLAN 11 Mbps [Feeney2001]  Transmitting  = 1 (normalized)‏  Receiving  R = 2/3  Sensing  S = 0.82  R  Possible power saving policy  Case 1 Energy spent during NAV phase is negligible (  0 =0)  Case 2 Energy spent during NAV phase is not negligible (  0 =0.5  S )  Case 3 Regular sensing is performed during NAV phase (  0 =  S )

12. WPMC'04 Abano Terme, Padova (Italy) 12-15 September 2004 0 10 20 30 40 50 60 70 80 90 100 Normalized Lifetime Number of stations Basic Access RTS/CTS  0 = 0  0 =  S  0 = 1/2  S 0.2 0.4 0.6 Results: node lifetime  Normalized Lifetime  Minimum theoretical energy per pck over Average energy per pck  RTS/CTS outperforms Basic Access mode   0 =0 leads to large gain in nodes lifetime  Gain rapidly fades for  0  1/2  S

13. WPMC'04 Abano Terme, Padova (Italy) 12-15 September 2004 0 10 20 30 40 50 60 70 80 90 100 Payload [bits] Number of stations Energy-based threshold Throughput-base threshold [1] 2000 4000 6000  0 = 0  0 =1/2  S 0 =S0 =S Basic Access-RTS/CTS threshold  Energy vs Throughput perspective  With  0  1/2  S payload threshold is lower than in Throughput-base case  Threshold shows less sensitivity to the number of nodes in the network With more than 20 nodes, the threshold remains almost const  Threshold increases as  0 gets close to  S Payload threshold after which RTS/CTS outperforms Basic Access [1] Bianchi2000

14. WPMC'04 Abano Terme, Padova (Italy) 12-15 September 2004 Conclusions  Complete statistical description for energy consumption  Ad-hoc network with saturated IEEE 802.11 nodes  Model allows for some interesting insights  Channel sensing during backoff has a relevant energetic cost  Switching to low-power mode during NAV can potentially save energy, but only for  0 <<  S  Payload length after which RTS/CTS outperforms Basic Access is lower for Energy-base than for Throughput-base perspective  Energy-based Threshold is less sensitive to the number of nodes in the network than Throughput-based Threshold

15. WPMC'04 Abano Terme, Padova (Italy) 12-15 September 2004 Department of Information Engineering University of Padova, ITALY Mathematical Analysis of IEEE 802.11 Energy Efficiency {andrea.zanella, depe}@dei.unipd.it Andrea Zanella, Francesco De Pellegrini WPMC 2004, 12-15 September 2004 Questions?

16. WPMC'04 Abano Terme, Padova (Italy) 12-15 September 2004 Extra Slides… Spare Slides

17. WPMC'04 Abano Terme, Padova (Italy) 12-15 September 2004 Medium Access Control (MAC)‏  CSMA: Carrier Sensing Multiple Access  (Exponential) Backoff stage Choose a random number in the backoff window If the channel is sensed idle, then countdown by 1 for each slot If the channel is busy then freeze the countdown until the channel becomes idle again for at least a DIFS  When the countdown is over transmit the packet If no ACK is returned within a SIFS, a collision has occurred Double backoff window and re-enter the backoff stage Otherwise the transmission was successfull Reset the backoff window and enter the backoff stage for the next packet

18. WPMC'04 Abano Terme, Padova (Italy) 12-15 September 2004 Collision Avoidance  Basic Access  Transmit data packet  RTS/CTS access  Try to reserve the channel before transmission Send a very short Request To Send (RTS) packet Receiver replies with a very short Clear To Send (CTS) packet Stations that get RTS or CTS packets avoid transmissions in the successive time interval (setting the NAV)‏

19. WPMC'04 Abano Terme, Padova (Italy) 12-15 September 2004 Detailing E B : backoff strategy  E B : Energy spent in backoff  Backoff strategy  S(i) : backoff stage after i successive collisions S(i) = min(i,m)‏  CW i : i-th backoff window CW i =CW 0 2 S(i) -1  x i : i-th backoff counter x i =random{0,1,...,CW i }  Countdown x i tick periods then retransmit the packet

20. WPMC'04 Abano Terme, Padova (Italy) 12-15 September 2004 Tick period (1/2)‏  Tick Period  time between two successive decrements (tick) of the backoff countdown process Idle channel countdown 1 per time slot Busy channel (valid or collided packet on the air)‏ freeze until the channel returns idle for a DIFS, then resume countdown  during a tick period a node can wait (idle channel)‏ receive valid packet intended for that node discard valid packet for other destinations listen collided transmission on the channel Idle Channel Busy Channel

21. WPMC'04 Abano Terme, Padova (Italy) 12-15 September 2004 0 20 40 60 80 100 120 140 160 180 2000 P [ E > e ] Normalized energy: e=E / min{E} Basic Access  0 =  S 10 -3 10 -2 10 -1 Results: complementary cdf of E  Energy actually spent for a packet transmission is many times the theoretical minimum  Jointly using RTS/CTS and smart sensing strategy drastically reduces energy costs 10 -4 10 0 Basic Access  0 = 0 RTS/CTS  0 =  S RTS/CTS  0 = 0