Model-Driven Energy-Aware Rate Adaptation M. Owais Khan, Vacha Dave, Yi-Chao Chen, Oliver Jensen, Lili Qiu, Apurv Bhartia, Swati Rallapalli The University of Texas at Austin MobiHoc.
Download ReportTranscript Model-Driven Energy-Aware Rate Adaptation M. Owais Khan, Vacha Dave, Yi-Chao Chen, Oliver Jensen, Lili Qiu, Apurv Bhartia, Swati Rallapalli The University of Texas at Austin MobiHoc.
Model-Driven Energy-Aware Rate Adaptation M. Owais Khan, Vacha Dave, Yi-Chao Chen, Oliver Jensen, Lili Qiu, Apurv Bhartia, Swati Rallapalli The University of Texas at Austin MobiHoc 2013, Bangalore, India 1 Motivation • Multi-antenna devices are becoming common • Offer diverse rate choices – # of antennas, modulation, coding, # of streams • Rate adaptation – beaten to death problem? • Large capacity gain, but significantly more energy! Mode Intel TX Intel Rx Single Antenna 1.28 W 0.94 W Two Antennas 1.99 W 1.27 W RateThree adaptation needs Antennas 2.10 W to energy-aware! 1.60 W What’s the big deal? • Fixed antenna systems are fairly simple 𝑒𝑛𝑒𝑟𝑔𝑦𝑐𝑜𝑛𝑠𝑢𝑚𝑒𝑑 ∝ 𝐸𝑇𝑇 • Energy-aware rate adaptation becomes simple Highest rate Lowest ETT Minimum energy! • Can this be applied to MIMO as well? 𝑒𝑛𝑒𝑟𝑔𝑦𝑚𝑢𝑙𝑡𝑖−𝑎𝑛𝑡𝑒𝑛𝑛𝑎𝑠 ≫ 𝑒𝑛𝑒𝑟𝑔𝑦𝑠𝑖𝑛𝑔𝑙𝑒−𝑎𝑛𝑡𝑒𝑛𝑛𝑎 – Additional hardware and RF chains – But multiple data streams reduces transmission time! Energy vs. Tx time: the trade-off % decrease required in tx time Reduce time by 68%! 70 3x3 Ant 2x2 Ant 65 60 55 Reduce time by 50%! 50 45 40 35 0 0.2 0.4 0.6 0.8 1 1.2 1.4 Single antenna tx time (ms) • Exact rate and # of antennas depend on multiple factors – Channel condition, wireless card and frame size 1. No single setting to minimize energy 2. Single antenna ≠ minimum energy Hence, our work! Understand how energy consumption in these devices relates to these factors Design an energy-aware rate adaptation scheme that can minimize energy! Contributions • Extensive power measurements for multiple 802.11n wireless adapters • Derive energy model based on power measurements • Propose an energy-aware rate adaptation scheme • Evaluate using simulation and testbed experiments Why Model-Driven? • Why not probing? – Slow given the large search space w/ MIMO – Hard to accurately measure the power of probe frames • Model-driven – Estimate power consumption for each rate under the current channel condition – Directly select the one w/ lowest power Power Measurement Setup • Multiple wireless cards – Intel 5300N series – Atheros 11n – Windows mobile smartphone • Monsoon power monitor – One reading/μs – Maximum power value every 200μs Power Measurement Methodology • Measurements at both transmitter and receiver • Different configurations – Frame size (250-1500 bytes) – # of antennas – 802.11n compliant data rates Atheros Energy Measurements 1.2 0.6 0.5 0.8 Energy (J) Energy (J) 1 0.6 0.4 3 Ant 2 Ant 1 Ant 0.2 0 0 0.5 1 1.5 0.4 0.3 0.2 3 Ant 2 Ant 1 Ant 0.1 0 2 2.5 0 0.2 0.4 0.6 0.8 1 Transmission Time (S) Transmission Time (S) Atheros Wi-Fi transmitter Atheros Wi-Fi receiver 𝑬𝒏𝒆𝒓𝒈𝒚𝒄𝒐𝒏𝒔𝒖𝒎𝒆𝒅 ∝ 𝑬𝑻𝑻 Slope of the line depends on # of antennas 3 3.5 2.5 3 2 2.5 Energy (J) Energy (J) Intel Energy Measurements 1.5 1 0 0 0.5 1 1.5 1.5 3 Ant,1 Streams 3 Ant,2 Streams 3 Ant,3 Streams 2 Ant,1 Streams 2 Ant,2 Streams 1 Ant,1 Streams 1 3 Ant,3 Streams 2 Ant,2 Streams 1 Ant,1 Streams 0.5 2 0.5 2 2.5 Transmission Time (S) Intel Wi-Fi transmitter 0 0 0.5 1 1.5 2 Transmission Time (S) Intel Wi-Fi receiver 𝑬𝒏𝒆𝒓𝒈𝒚𝒄𝒐𝒏𝒔𝒖𝒎𝒆𝒅 ∝ 𝑬𝑻𝑻 Slope of the line depends on # of antennas 2.5 Measurement-Driven Energy Model • Use least-square fitting to develop energy models 𝐸𝑡𝑥 = 𝐴 × 𝐸𝑇𝑇 + 𝐵 𝐸𝑟𝑥 = 𝐶 × 𝐸𝑇𝑇 + 𝐷 where 𝐴, 𝐵, 𝐶, 𝐷 vary for different wireless cards Intel Atheros 𝐴 0.24 × 𝑛𝑡𝑥 + 0.425 × 𝑀𝐼𝑀𝑂 + 1.02 0.38 × 𝑛𝑡𝑥 + 0.108 𝐵 0.045 × 𝑛𝑡𝑥 + 0.108 0.040 × 𝑛𝑡𝑥 + 0.062 𝐶 0.30 × 𝑛𝑟𝑥 + 0.61 0.142 × 𝑛𝑟𝑥 + 0.30 𝐷 0.064 × 𝑛𝑟𝑥 + 0.167 0.048 × 𝑛𝑟𝑥 + 0.106 Validating the model 𝑥 − 𝑥′ 𝑀𝐴𝑃𝐸 = 𝑚𝑒𝑎𝑛( ) 𝑥 𝑥 : actual energy consumption 𝑥 ′ : estimated energy consumption Card Atheros Intel Phone Transmission 3.4% 0.65% 4.9% Reception 1.3% 1.4% 3.6% Error is consistently below 5%! Energy Aware Rate Adaptation Select rate for next transmission that minimized energy! Channel State Information (CSI) • Specifies amplitude and phase between tx-rx pair – Measured for all subcarriers using preamble – Reported once per received frame • pp-SNR can be calculated as: 𝑀𝑀𝑆𝐸 𝑆𝑁𝑅𝑚 𝐸𝑠 = 𝑁𝑡 𝑁0 [𝐻𝐻 𝐻 + 1 𝐸𝑠 𝑁𝑡 𝑁0 −1 𝐼]−1 Compute loss rate • Map pp-SNR to un-coded BER using known relationship BERuncoded £ c1 ´Q ( c2 ´ SNR • Convert un-coded BER to coded BER BERcoded ( r ) = ¥ å a d ×Pd ( r ) d=d free • Calculate frame error rate (FER) FER =1- (1- BERcoded )L ) Estimate energy consumption • AP or back-end server keeps table of energy models – Account for most commonly used Wi-Fi cards • Get the make/model of the Wi-Fi card – Explicit feedback or passive detection • Compute ETT based on frame loss rate (FER) 𝑠𝑖𝑧𝑒 1 𝐸𝑇𝑇 = 𝑟𝑎𝑡𝑒 (1 − 𝐹𝐸𝑅) • Select the MCS with the minimum energy – Different variations of schemes are possible • Partial packet recovery (PPR) support – Only the ETT calculation changes (ref. paper) Putting it all together Measure CSI Calculate pp-SNR Calculate estimated loss rate Compute ETT Select rate minimizes energy! Evaluation • Trace-driven simulator – Static and mobile channel traces using Intel 5300 – Written in python • Testbed – Uses the Intel 5300 card – Iwlwifi driver is modified to support rate adaptation Simulation Methodology • Developed in Python using real CSI traces • Different schemes are supported – – – – – Sample Rate with MIMO Effective SNR Maximum throughput Minimum energy Minimum Energy with throughput constraint Intel Transmitter (Static) 14%-24% 17%-31% 1-10% 1-2% 10%-22% 51 66 46 Throughput (Mbps) 76 Energy (nJ/bit) 56 46 36 26 16 1-19% 0-1% 1-2% 41 36 31 26 21 16 11 6 6 Trace 1 Trace 2 Trace 3 Trace 1 Trace 2 Trace 3 Minimum Energy provides significant power savings 21 Intel Receiver (Static) 25-35% 26-42% 1-10% 1-4% 10%-26% 86 76 76 Throughput (Mbps) 86 Energy (nJ/bit) 66 56 46 36 26 46 36 26 6 6 Trace 3 1-5% 56 16 Trace 2 0-1% 66 16 Trace 1 1-23% Trace 1 Trace 2 Trace 3 22 Intel Receiver (Mobile) 29-31% 34-40% 1-16% 2-6% 106 96 86 76 66 56 46 36 26 16 6 15-19% 2-13% 0-2% 3-6% Energy (nJ/bit) Throughput (Mbps) 41 36 31 26 21 16 11 6 Trace 1 Trace 2 Trace 3 Trace 1 Trace 2 Trace 3 23 Intel Receiver with PPR 76 Minimum Energy Maximum Tput w/ 80% Tput constraint w/ 60% Tput constraint 26% to 28% ~9% Trace 1 Trace 2 Throughput (Mbps) Energy (nJ/bit) 66 56 46 36 26 16 6 Trace 3 56 51 46 41 36 31 26 21 16 11 6 26% to 29% Trace 1 ~9% Trace 2 Trace 3 Minimum Energy provides significant power savings 24 Always use single-antenna systems? 1 antenna 2 antennas 3 antennas 1.0 0.8 0.6 0.4 0.2 0.0 1000 2000 3000 4000 Packet size (Bytes) 5000 Single Antenna performance decreases with increasing packet size Testbed • Implemented scheme on Intel Wi-Fi link 5300 driver – Used tool in [Halperin10] to extract CSI from driver • Static channel – 200 UDP Packet of 1000 bytes each transmitted – Results averaged over 10 runs • Mobile channel – Receiver moves away from transmitter at walking speed – Results averaged over 5 runs Static Channel Minimum Energy w/ 80% Tput constraint 76 19% 6% 28% 16% 51 Throughput (Mbps) 66 Energy (nJ/bit) Maximum Tput w/ 60% Tput constraint 56 46 36 26 16 24% 11% 22% 2% 46 41 36 31 26 21 16 11 6 Transmitter Receiver 6 Transmitter Receiver 27 Mobile Channel Energy savings do not degrade with the channel! Related Work • Models based on data size [Carvahlo04], empirical study [Bala09] • Neither considers effects of multiple antennas, data rates, tx power • Study power consumption under different parameters[Halperin10] • Do not develop energy model Energy measurement and Models • Extensively studied [Bicket05, Holland01, Sadeghi02, Wong06, etc.] • None of these schemes consider minimizing energy • Energy based rate adaptation [Li12] • Limited effectiveness of probing-based approach Rate Adaptation • Power Saving Mode Optimization [Napman10, Sleepwell11, E-mili11] • Complementary to our work Power Savings 29 Conclusion • Collect and analyze extensive power measurements – Derive simple energy models for transmission/reception • Develop model-driven energy-aware rate adaptation scheme • Experimentally show significant energy savings possible – 14-37% over existing approaches – PPR extensions can be even better Questions ??? Thank You! [email protected]