ARC Special Research Centre for Ultra-Broadband Information Networks

University of Melbourne

Energy Consumption of the Internet

Jayant BaligaJayant Baliga

Robert W. Ayre, Kerry Hinton, Wayne V. Sorin, Rodney S. TuckerRobert W. Ayre, Kerry Hinton, Wayne V. Sorin, Rodney S. Tucker

Metro/Edge

Network

Core Network IPTV Network

Data

CenterAccess

Network

• Greenhouse Impact

• Energy-limited capacity bottlenecks (“hot spots”)

• Operational Expenditure (OPEX)

Why is Energy Important?

Energy and the Internet

•• A model to estimate energy consumption of the InternetA model to estimate energy consumption of the Internet

– Core, metro, and access networks

•• Where does the energy go? Where does the energy go?

•• What happens as traffic grows?What happens as traffic grows?

•• Towards an energy efficient InternetTowards an energy efficient Internet

Calculating Energy Consumption of the Internet

EthernetSwitch

OLTSplitter

Cabinet

Access NetworkMetro/Edge NetworkCore Network

Edge Routers

OLT

SwitchONU

Cabinet

PtP

FTTN

PON

DSL

Fiber

EDFA

Core Router

OXC

IPTV Network

Storage

Server

Server

Storage

Data Center

DSLAM

DSLAM

Cu

Fiber

Cu

Fiber

Broadband Network Gateways

Tier 1 Full-Service ISP Network

Network Energy Model

Peak access rate sold to userM =

Capacity per user

Oversubscription~ 2.5 Mb/s in 2008

~ 0.1 Mb/s in 2008

M = 25

•• Choose an access data rate (capacity per user)Choose an access data rate (capacity per user)

•• Carry out paper design of networkCarry out paper design of network

•• Calculate the power consumed by the network per userCalculate the power consumed by the network per user

•• Repeat for all access rates Repeat for all access rates

Estimating Energy Consumption

1500

Po

wer

(W/u

ser)

Baliga et al., 2008

0

20

25

5

Peak Access Rate (Mb/s)

15

50 100

10

Total

Routers

Access (PON)Transmission Links

Today’s Internet

(~ 2.5 Mb/s)

2008 Technology10 core hops

Oversubscription = 25

% o

f E

lectr

icit

y S

up

ply

0.5

1.0

Power Consumption of IP Network

Wave7 ONT-G1000i

Splitter GPON

PtP

Cabinet

Edge Node

Cisco

6513

Hitachi 1220

NEC AM3160

Cisco 4503

NEC VF200F6

TC Communications TC3300

Axxcelera

ExcelMax

Cabinet

Access N/W

FTTN VDSL2

Cu

Fiber

Fiber

Cisco uBR10012

RF Gateway Cisco DCP3000

HFC

Cu

RF Amp

Node

WiMAX

Axxcelera

ExcelMax BTS

Fiber

Hitachi 1220

Splitter

Power Consumption in Access Networks

Power Consumption in Access Networks

30

Peak Access Rate (Mb/s)

Po

wer

Per

User

(W)

1 1000

0

PtP

PON

100

20

10

10

32 Customers M = 32

M = 1

M = 1

M = 1

PON FTTH is PON FTTH is ““greenestgreenest””

FTTN

WiMAX

M= 10

20 users per sector

HFC

M= 10M= 1

M = Oversubscription

•• Access network dominates at low ratesAccess network dominates at low rates

•• Network routers dominate at higher ratesNetwork routers dominate at higher rates

•• Transmission Links consume a small percentage of the total Transmission Links consume a small percentage of the total

powerpower

•• Possible approaches to controlling growth in energy Possible approaches to controlling growth in energy

consumption:consumption:– Improve electronic technology

– New architectures (Optical bypass)

– New protocols (“low energy” states)

Some Observations

Based on G. Epps, CISCO, 2006

1

10

100

1000

10000

1994 1996 1998 2000 2002 2004 2006

X 1

993 P

erf

orm

an

ce

2008

CRS-1

(~1.3 Tb/s,13.6 kW/rack)

12416

(~0.3 Tb/s/rack)

12016

(~80 Gb/s/rack)

Year

Moore’s law x 2/18 m

Router capacity x 2.5/18 m

CMOS energy efficiency x 1.65/18 m

2010

30% p.a.

Router energy efficiency

improving at 20% p.a.

Neilson, JSTQE 2006

Router Capacity Growth

•• Router energy consumption is reaching the limits of air coolingRouter energy consumption is reaching the limits of air cooling

– Cisco CRS-1 (largest core router available)

Neilson, 2006 & Deutche Telekom, 2007

Per Rack

X2 every

18 months

The Energy Bottleneck

15000

Po

wer

(W/u

ser)

% o

f E

lectr

icit

y S

up

ply

Baliga et al., 2008

0

20

25

5

Peak Access Rate (Mb/s)

15

10

2008 Technology

Effect of Efficiency Gains?

Routers

Access (PON)Transmission Links

0.5

1.0

Total

50 100

To

tal

Po

wer

(W/u

ser)

Baliga et al., JLT 2009

0

20

25

5

Peak Access Rate (Mb/s)

15

10

Overall Technology Efficiency

Improvement Rate

= 0% p.a 5% p.a

Traffic Growth rate = 40% p.a

20% p.aTarget

10% p.a

% o

f E

lectr

icit

y S

up

ply

Improvements in Technology Efficiency

1500 50 100

0.5

1.0

0

No bypass

Peak Access Rate (Mb/s)1 10 100 300

To

tal

Po

we

r P

er

Us

er

(W)

0

10

20

5

0

% o

f E

lec

tric

ity C

on

su

mp

tio

n

0.25

0.5

0.75

Bypass

Access

Efficiency Improvement Rate = 10% p.a.

15

1.0

Destination

Bypass Bypass

Router

X-connect

Baliga, et al, JLT 2009

Source

Improvements in Network Architecture

Optical BypassOptical Bypass

Summary – Where to in the Future?

•• Energy consumption of the Internet is small (0.4%), but will Energy consumption of the Internet is small (0.4%), but will

approach 1% in the futureapproach 1% in the future

•• Internet energy consumption dominated byInternet energy consumption dominated by

– Access network today

– Core network in the future

•• A multiA multi--disciplinary approach is required to build a green disciplinary approach is required to build a green

Internet:Internet:

– Improved efficiency in electronic and photonic devices

– Improved architectures