Are Optical Networks Green? Rod Tucker University of Melbourne

Are Optical Networks Green? Rod Tucker University of Melbourne

Historical Perspective Transport Systems Energy/Bit/1000-km (J) 10 6 10 3 1 10-3 10-6 10-9 Marconi Trans-Atlantic Fessenden Trans-Atlantic First Trans-Atlantic Newhaven - Azores NY - Paris Key West - Havana ~20% p.a. improvement TAT-1 TAT-3 TAT-5 TAT-9 TAT-12/13 TAT-8 WDM terrestrial Wireless Telegraphy Coax Optical + Regen. Optical + EDFA TAT-10 TAT-11 10-12 1860 1880 1900 1920 1940 1960 1980 2000 2020 Year 2040

Summary Top-down estimate of energy consumption of the Internet - Projections of current trends - es and routers - Optical transport - Access network Bottom-up estimate - Based on theoretical and practical lower bounds - Transport energy - ing energy - Network energy How to build an energy-efficient network

Network Energy Model Core Network Tier 1 Network Metro/Edge Network Access Network Core Router Fiber Broadband Network Gateways Ethernet DSLAM Cu DSL Edge Routers OLT Cabinet Splitter Fiber FTTP OLT Cabinet Cu FTTN ONU DSLAM Server Storage Data Center Server Storage Video Distribution Network

Power Consumption in Access Networks 30 M= Oversubscription Power Per User (W) 20 10 Wireless M = 1 M= 1 M= 10 M= 10 HFC PON Fiber to the Node M = 1 2010 Technology M = 1 Point to Point Optical M ~10 0 1 10 100 1000 Peak Access Rate (Mb/s) FTTP is greenest

Energy per Bit in Network Devices 1000 2010 2010 Energy per bit (nj) 100 10 1 0.1 Wavelength Sub-wavelength Estimated 0.01 MEMS OXC Optical Amp PIC Tx/Rx Set-top Box: Discrete Tx/Rx PoS Tx/Rx 1000 nj/b Ethernet Core Router PON ONU (10 Mb/s) IPTV Server HD IPTV: 10,000 nj/b

Energy Efficiency Improvements with Time Transport Energy/Bit/1000-km (J) 10-3 10-6 10-9 10-12 20 % p.a. TAT-1 TAT-3 TAT-5 TAT-9 TAT-12/13 TAT-8 WDM terrestrial TAT-10 TAT-11 1920 1940 1960 1980 2000 2020 Year 2040 20% p.a. efficiency improvement in routers (Neilson, JSTQE, 2006) 13% p.a. efficiency improvement in routers (Tamm, BLTJ, 2010) 15% p.a. efficiency improvements in transport (Han, IEEE Comms. Mag. 2010)?

Network Energy Per Bit 100 Total using 2010 Technology 20 hops Energy per bit (μj) 10 1.0 0.1 Routers and switches Transport PON Total (20% p.a. improvements) 40% Access Rate Growth 0.01 2.5 10 100 250 2010 Peak Access Rate (Mb/s) 2020

Global Network Energy Consumption Global Network Power Consumption (W) 10 12 0% p.a. 2010 Global efficiency electricity supply improvement 10 11 Total 10 10 10 9 10 8 Routers and switches Transport PON 40% Access Rate Growth 10% Growth in user numbers 10 7 2.5 10 100 250 2010 Peak Access Rate (Mb/s) 2020 20% p.a. efficiency improvement 1.2 Billion Users

Lower Bounds on Network Energy Consumption Top-down estimate of energy consumption of the Internet - Projections of current trends - es and routers - Optical transport - Access network Bottom-up estimate - Theoretical and practical lower bounds on energy - Transport energy - ing energy - Network energy How can we build a more energy-efficient network

Lower Bound on Optical Transport Energy Stage 1 Stage m P TX P A P A P RX TX α g α g RX L stage Amplifier Energy TX/RX Energy E E P A AMP min = Br TX / RX min P α L 2 stage SNR m e h + bit B TX min RX min = Dominates B r P r ν Bit Rate

Optical Transmitters P MUX P driver Data In MUX Driver V mod I laser Z mod P CW P 1 V laser I mod Modulator Z mod C mod = or 50 Ω Lumped modulator Distributed modulator 1 E = C V 2 E = 2 mod mod V 50B 2 mod r C mod 1 25B < 50Ω r (Tom Koch)

Minimum Amplifier Energy per Bit Minimum Amplifier Energy, E AMP-min (pj/bit) 1.0 0.8 0.6 0.4 0.2 Shannon bound for SE = 1 b/s/hz TX/RX ( ~2020) OOK DBPSK 0 0 20 40 60 80 100 120 140 160 Amplifier Spacing, L stage (km) Amplifiers (Theoretical Lower Bound)

Energy per Bit per 1000-km Energy/Bit/1000-km (J) 10-3 10-6 10-9 TAT-1 TAT-3 TAT-5 TAT-9 TAT-12/13 TAT-8 WDM terrestrial 10-12 Minimum TX/RX/amplifier energy TAT-10 TAT-11 ~ x 10 3 1920 1940 1960 1980 2000 2020 2040 Year

Networks Stage 1 Transport Transport Stage s User Ports Access Network Access Network Clos

Optical es Optical Optical Inputs O/E/O O/E/O O/E/O Optical O/E/O O/E/O O/E/O Optical Outputs O/E O/E O/E Electronic E/O E/O E/O

O/E Interfaces Electronic Demultiplexing O/E Electronic DEMUX Electronic (a) TDM Optical Demultiplexing Optical O/E O/E O/E Electronic (b) Advanced Modulation Formats DQPSK, OFDM, etc. Electronic (c) Transport 1 1000 Gb/s ing and Processing 0.1 10 Gb/s

Optical Technologies Electro-optic (O-E) SOA gate arrays AWG-based wavelength-routed switches CMOS

Packet ing E AMP E control τ p τ b Optical Array Inputs Outputs E bit = E + AMP E control N b τ τ p b ~ 10 4 for IP packets

Energies per Bit O/E Converters MUX/DEMUX

Global Network Network Energy per bit (J) 10-5 10-6 10-7 10-8 10-9 10-10 10-11 Transport X X Transport Routers and es es Equipment Data Current Trends (Moore s Law) Lower Bound Limits X X ~ x 10 4 difference 10-12 1 2.5 10 100 250 2010 2020 Peak Access Rate (Mb/s) 1000

Loss/Efficiencies and Energy Overheads Laser efficiency, system penalties, system margins, etc E function E = η E Total min function E overheads E E min Total = 10 10 2 4 Loss, Inefficiencies E min Key Function Overheads Subsystem Management and control, interconnects, power supplies, etc. Key Conclusion: Minimizing E min is not necessarily the best strategy for minimizing E Total

Power Consumption in Electronic Routers Security, 25% Route processors, 8% Forwarding plane, 25% Electronics I/O, 11% fabric, 15% Buffers, 5% Potentially Optical, 20% Source: G. Epps, Cisco Control plane, 12%

How to Build an Energy-Efficient Efficient Network Focus on the access network and Customer Premises Equipment Fiber to the Premises, sleep modes set-top boxes, IP-enabled HDTV s, etc. Reduce energy losses and overheads E.g., high-efficiency modulator drivers, EDFA pumps, ancillary circuitry All-optical transmission and switching may be of little help Reduce number of network hops OADM s, Layer 2 versus Layer 3, dedicated IPTV distribution Operate network at high utilization QoS issues, traffic engineering, sleep modes in the core

Data by Mail: Data by Mail vs. Data by the Internet Cargo Jet 10 8 32-GB USB drives Melbourne 5x10 6 Kg CO 2 (~24 hours) San Diego Data by Internet: 3x10 9 GB The Internet 1000 Gb/s for 3x10 7 seconds 2x10 7 Kg CO 2 (1 year)

(www.greentouch.org) Global consortium, launched January 12 Goal: x10 3 improvement in energy efficient of the network by 2020 Members: - Bell Labs,Telifonica, Huawei, AT&T, China Mobile, Freescale Semiconductor, Swisscom, Portugal Telecom, SAIT, INRIA, IMEC - University of Melbourne (IBES), MIT, Stanford - More to follow Outcomes: - Reductions in carbon footprint and operating cost - Collaboration between leading experts from around the world - Opportunities to bring innovative new ideas and products to market