Energy and Cost Efficient Ultra-High Capacity Wireless Access Sibel Tombaz, Västberg, Jens Zander Sibel Tombaz, Anders Västberg, Jens Zander Wireless@KTH, Royal Institute of Technology, Stockholm, Sweden eWIN • National project – within the Strategic Research Area ”ICT – The Next Generation” GreenNet Workshop IEEE VTC2011 Budapest 2 1 Data consumption trends Tablet Computers Voice-dominated consumption Super Phones Smart Phones Feature Phones Current values Web-brow. 40MB Spotify-like 60MB YouTube 200MB Predictions Web-brow. 60 MB per hour Spotify-like 60MB YouTube 400MB Data-dominated consumption 3 GreenNet Workshop IEEE VTC2011 Budapest Why Energy Efficiency ? Traffic, $ Traffic, cost & energy consumption The Revenue Gap traffic vs. Telcos’ revenues decoupling Telcos’ revenues Voice dominant Data dominant Time • Energy costs grow even faster than traffic volumes • Cost is a more effective drive than environmental concerns GreenNet Workshop IEEE VTC2011 Budapest 4 2 Solutions • Short term – efficient implementation – Improved efficiency of amplifiers/electronics/antennas/SP • Medium term – exploit existing standards – Provide capacity when its needed: Idle power & Transmit Power Management – Provide capacity where its needed: HETNETs – infrastructure & energy spent close to the demand. GreenNet Workshop IEEE VTC2011 Budapest 5 HETNET deployment (Klas Johansson, ”Cost Effective Deployment Strategies for Heterogeneous Wireless Networks”, Doctoral Thesis, KTH 2007) GreenNet Workshop IEEE VTC2011 Budapest 6 3 Solutions • Short term – efficient implementation – Improved efficiency of amplifiers/electronics/antennas/SP • Medium term – exploit existing standards – Provide capacity when its needed: Idle power & Transmit Power Management – Provide capacity where its needed: HETNETs – infrastructure & energy spent close to the demand. • Long term – Clean slate design – – – – Revisit constraints: spectrum, infrastructure cost, energy cost Clean slate design: How would an energy efficient design look like ? What would a design minimizing the TOTAL cost look like ? Are we “way off” today 7 GreenNet Workshop IEEE VTC2011 Budapest Key Design Constraints Energy C A: Current designs B: Femto/Picocell design C:” Green” design E D: Total cost minimization – abundant spectrum D Infra cost B C tot = C infra + C energy GreenNet Workshop IEEE VTC2011 Budapest A E: Total cost min – less spectrum available Spectrum + C spectrum 8 4 Outline – Clean slate design • Objectives : – Modeling energy consumption in Wireless Networks – Simple model to allow design space exploration – Characterizing trade-offs future architectures • Two model architectures – “Green architecture” – providing the required service at minimal energy cost – Architectures for total cost minimization • Conclusions GreenNet Workshop IEEE VTC2011 Budapest 9 Energy consumption modelling • Assumptions: – – – • Homogenous network constant traffic over cell size Real network – composed of (tiers of) homogeneous “islands” Given throughput requirement in area Rtot Power consumption Ρtot = N BS [aPtx + bradio + bbackhaul ] + d + yRtot “Efficiency” Proportional to #base stations •Idle power, •Control msg transmission • backhaul .. GreenNet Workshop IEEE VTC2011 Budapest Independent of #base stations: •core network 10 5 HETNET deployment (Klas Johansson, ”Cost Effective Deployment Strategies for Heterogeneous Wireless Networks”, Doctoral Thesis, KTH 2007) 11 GreenNet Workshop IEEE VTC2011 Budapest Energy consumption modelling (2) • Spectrum-Infrastructure Cost-Power Trade-off (Shannon Bound) Prx (d ) = • c' GPtx dα Average spectral efficiency N W Ρtot = N BS a o cG R NW α Ptx = 2W − 1 0 Rcell cG S = Rt tot 2 N BSW − 1 Α πN BS GreenNet Workshop IEEE VTC2011 Budapest R W α /2 Rtot +d / A + bradio + bbackhaul + y N BS 12 6 “Green” Architecture • If power/energy is the dominant constraint; lim Pc ( N BS ) →∝ α >2 N BS →∝ 3000 W=1.25 Mhz W=3.5 Mhz W=5Mhz W=10Mhz Area Power Consumption (Watt/km 2) 2500 W=20Mhz 2000 1500 1000 500 0 0 5 10 15 20 25 NBS 30 35 40 45 50 There is always a non-null and finite that minimizes the areapower consumption. 13 GreenNet Workshop IEEE VTC2011 Budapest Green Architecture • Idle power PA - efficiency 2000 3000 bradio=0 1800 bradio=100 1600 bradio=300 bradio=1000 2000 Rtot=20Mbps/km2 1500 Rtot=5Mbps/km2 1000 500 Area Power Consumption (Watt/km 2) Area Power Consumption (Watt/km 2) 2500 bradio=a a=0 a=5 a=22 a=50 a=100 Rtot=20Mbps/km2 1400 1200 1000 800 Rtot=5Mbps/km2 600 400 200 0 0 5 10 15 20 25 NBS 30 35 40 45 50 0 0 5 10 15 20 25 NBS 30 35 40 45 50 Optimistic model: a(S)= const GreenNet Workshop IEEE VTC2011 Budapest 14 7 Deployment for Minimal Total Cost C tot = C infra + C energy N W Cost = co * N BS + c1 N BS a o cG Rt tot 2 N BSW − 1 Α πN BS Infrastructure α /2 + C spectrum Rtot + d + c2W + bradio + bbackhaul + y N BS Energy Spectrum 15 GreenNet Workshop IEEE VTC2011 Budapest Deployment for Minimal Total Cost 2000 Energy Cost Infra Structure Cost Spectrum Cost Total Cost 1800 1600 Cost (kEuro) 1400 1200 1000 800 600 400 200 0 • • 0 5 10 15 20 25 NBS 30 35 40 45 50 Minimum total cost now occurs at a much lower number of base stations than in the energy-only minimization. Spectrum cost constant – provides only a level shift of the total cost; GreenNet Workshop IEEE VTC2011 Budapest 16 8 Increasing infrastructure cost 3000 Total Network Cost (kEuro) 2500 c2=0.05MEuro/BS/year 2000 1500 1000 500 infra cost increases c2=0.02MEuro/BS/year Rtot 0 • • 0 5 10 15 20 25 NBS 30 35 40 45 50 Total cost increases – same effect as idle power/backhaul cost Optimal number of base station is not that much affected. 17 GreenNet Workshop IEEE VTC2011 Budapest Spectrum cost impact 1400 Energy Cost Infra Structure Cost Spectrum Cost 1200 Total Cost 1000 kEuro 800 600 400 200 0 • 2 4 6 8 10 12 14 W [Mhz] 16 18 20 22 As the spectrum cost increases, optimum spectrum expenditure moves closer to the “energy asymptote” GreenNet Workshop IEEE VTC2011 Budapest 18 9 Energy price dependence 1.8 Optimum Base Station Density (# BS/km2) W=2 Mhz W=5 Mhz 1.6 W=10 Mhz 1.4 1.2 1 0.8 0.6 0.4 0.2 -1 10 • • 0 10 c 1-Cost of Unit Energy [Euro/kWh] 1 10 As the energy cost increase, distinct minimum moves closer to the “energy asymptote” where denser deployment is better. More available spectrum always decreases the optimum deployment density. GreenNet Workshop IEEE VTC2011 Budapest 19 Conclusions • Proposed simple Power/Energy consumption model allows for “clean slate” design space exploration Specific findings: • Increasing spectral efficiency further not really feasible – severe penalty in energy cost • Increase spectrum allowance significantly reduces the energy consumption (a “cost” - not a “must” issue) • There are unique optimal deployment densities that minimize energy and/or total cost • Optimum density – still fairly close to “Shannon Brick Wall” GreenNet Workshop IEEE VTC2011 Budapest 20 10 Read more & Interact ! wireless.kth.se GreenNet Workshop IEEE VTC2011 Budapest theunwiredpeople.com 21 Additional Slides 11 A new ballgame from “full coverage” to capacity (Klas Johansson, ”Cost Effective Deployment Strategies for Heterogeneous Wireless Networks”, Doctoral Thesis, KTH 2007) 23 GreenNet Workshop IEEE VTC2011 Budapest Cost drivers Greenfield deployment Cost structure, new sites 1,00 Transmission 0,80 Site buildout, installation and lease 0,60 Base station equipment, RNC, O&M and power 0,40 0,20 0,00 M acro M icro Pico WLAN (Klas Johansson, ”Cost Effective Deployment Strategies for Heterogeneous Wireless Networks”, Doctoral Thesis, Transmission KTH 2007) Radio equipment , O&M, power Sit e buildout , inst allat ion + lease 24 GreenNet Workshop IEEE 24 VTC2011 Budapest 12