Escolar Documentos
Profissional Documentos
Cultura Documentos
NETworks
ATNAC 2010 4
How much does energy cost
?
In the U.S.A.:
– A kWh is about $0.10 (US national average
billing rate)
– 1 TWh is about $100 million
In Italy:
– A kWh is about €0.085 (rate based on a
typical residential consumption of 2700
kWh for 1 year, under 3 kW)
– So, a TWh is €85 million (= ~$125 million)
ATNAC 2010 5
The Embodied Carbon
Footprint
Beyond the carbon footprint from direct
usage, we have also to take into account
the greenhouse gas emissions, needed to
produce devices (embodied carbon).
For example, in a PC:
– Carbon from usage (energy consumption):
4200 kWh = 2.1 tons of CO2 (5 years of
life)
– Embodied carbon: 2000 kWh = 1 ton of CO2
Embodied carbon
2002 0.11 0.43 0.53 Footprint from use
1.25%of total footprint
20
10
W
h
] m
C
ti[T
p
u
oE
rg
s
yn
e
0
2005 2010 2015 2020
Energy consumption estimation Year
for the European telcos’
network infrastructures in the “Business-As-Usual” (BAU) and
in the Eco sustainable (ECO) scenarios, and cumulative
energy savings between the two scenarios.
ATNAC 2010 10
… and the Future?
OPEX savings
4
BAU
ECO
3
2
n
ilo
$ b
0
2005 2010 2015 2020
OPEX estimation related to energyYear
costs for the European
telcos’ network infrastructures in the “Business-As-Usual”
(BAU) and in the Eco sustainable (ECO) scenarios, and
cumulative savings between the two scenarios.
18.12 4.53
ATNAC 2010 12
And the Reasons?
10000
To su p p o rt n e w
Router capacity TrafficLoad
x 2.5/18m x2/18m
g e n e ra tio n n e tw o rk
(Moore’slaw)
1000
ra p id ly g ro w in g CMOSenergyefficiency
te lco s a n d IS Ps n e e d :
m
n
a
c P
rfo
e
3
9
1
1
of devices,
– devices with
sophisticated Evolution from 1993 to 2010 of high-end IP
architectures able routers’ capacity (per rack) vs. traffic
to perform volumes (Moore’s law) and energy efficiency
increasingly in silicon technologies.
complex operations
in a scalable way. Source: R. Bolla, R. Bruschi, F. Davoli, F.
T h e so le in tro d u ctio n o f Cucchietti, “Energy Efficiency in the
n o ve llo w co n su m p tio n Future Internet: A Survey of Existing
Approaches and Trends in Energy-Aware Fixed
silico n te ch n o lo g ie s Network Infrastructures,” IEEE
ca n n o t cle a rly co p e
w ith su ch tre n d s, a nATNAC
d Communications Surveys & Tutorials, in
2010
b e e n o u g h fo r d ra w in g press 13
Decomposing the Energy
Consumption
Typical access, metro and core device density and energy requirements in
today’s typical networks deployed by telcos, and ensuing overall energy
requirements of access and metro/core networks.
ATNAC 2010 14
Decomposing the Energy
Consumption
ATNAC 2010 17
Re - engineering
R e -e n g in e e rin g a p p ro a ch e s a im a t:
– introducing and designing more
energy-efficient elements for
network device architectures
– suitably dimensioning and optimizing
the internal organization of devices
– reducing their intrinsic complexity
levels.
ATNAC 2010 18
Re - engineering
Energy -
Efficient Silicon
Adoption of pure optical switching
architectures:
– They can potentially provide terabits of
bandwidth at much lower power dissipation
than current network devices.
– But their widespread adoption is still hindered
by technological challenges: problems mainly
regard the limited number of ports and the
feasibility of suitable buffering schemes.
Decreasing feature sizes in semiconductor
technology have contributed to performance
gains:
– allowing higher clock frequencies
– designing improvements such as increased
parallelism.
– the same technology trends have also allowed
for a decrease in voltage
ATNAC 2010 that has reduced the19
power per byte transmitted by half every two
Re - engineering
Complexity
Reduction
Roberts proposed a radical new concept for
traffic lookup, which allows next-generation
routers forwarding traffic at the flow
levels.
– This approach certainly leads to a more scalable
and simple network device architecture with
respect to the current ones, which forward
traffic at the packet level.
Source: L. G. Roberts, “A Radical New Router”, IEEE
Spectrum, vol. 46, no. 7, pp. 34-39, July 2009.
Baldi and Ofek suggest a synchronous time-
based IP switching approach, which allows
synchronizing the operation of routers and
scheduling traffic in advance.
Source: M. Baldi, Y. Ofek, “Time for a “Greener” Internet,” Proc.
Green Communications Workshop in conjunction with IEEE ICC'09
(GreenComm09), Dresden, Germany, June 2009.
ATNAC 2010 20
Dynamic Adaptation
The dynamic adaptation of network/device
resources is designed to modulate
capacities of packet processing engines
and of network interfaces, to meet actual
traffic loads and requirements.
This can be performed by using two power-
aware capabilities, namely, dynamic
voltage scaling and idle logic, which both
allow the dynamic trade-off between
packet service performance and power
consumption.
ATNAC 2010 21
Dynamic Adaptation
Standard operations
Idle logic
ATNAC 2010 22
Dynamic Adaptation
IEEE 802 . 3 az
Green Ethernet (IEEE 802.3 az):
– First version: Adaptive Link Rate proposed
by Christensen and Nordman
Background:
10
5
0
10 100 1000 10000
Link speed (Mb/sec)
ATNAC 2010 23
Dynamic Adaptation
IEEE 802 . 3 az
Idea: switch the Ethernet link speed according to
incoming traffic volumes.
Effect: perform a link re-negotiation at each link
speed switch:
- proposal of a new mechanims for link speed
negotiation.
- need of interfacing Ethernet PHY chips with
frame Tx queues
Source: C. Gunaratne, K.
Christensen, S. Suen, B.
Nordman, “Reducing the
Energy Consumption of
ATNAC 2010Ethernet with an Adaptive24
Dynamic Adaptation
IEEE 802 . 3 az
Fin a lV e rsio n : b a se d o n th e “ lo w p o w e r id le ” co n ce p t,
p ro p o se d b y In te l.
– Idea: transmit data at the maximum speed, and put the
link to sleep when it is idle.
–
–
–
–
–
– Effect: LPI has two transitions for each packet (or block
of packets) : Link wake-up and sleep
• LPI can possibly be asynchronous (one direction
awake, the other asleep)
• Retraining ATNAC
can be2010
done via periodic on intervals 25
(if no packets are being sent)
Dynamic Adaptation
IEEE 802 . 3 az
ATNAC 2010 26
Dynamic Adaptation
SW Routers & ACPI
In PC-based devices, the Advanced Configuration
and Power Interface (ACPI) provides a
standardized interface between the hardware
and the software layers.
ACPI introduces two power saving mechanisms,
which can be individually employed and tuned
for each core:
– Power States (C-states)
• C0 is the active power state
• C1 through Cn are processor sleeping or
idle states (where the processor
consumes less power and dissipates less
heat).
– Performance States (P-states)
while in the C0 state, ACPI allows the
performance of the core to be tuned through P-
state transitions.
P-states allow to modify the operating energy 27
point of a processor /core
ATNAC by altering the
2010
Dynamic Adaptation
SW Routers & ACPI
The multi-core/cpu SW router architecture:
Tx Board
TxRings interfaces
High speed
interfaces
CPUs/cores
for application
services
Source: R. Bolla, R. Bruschi,
“PC-based Software Routers:
At least High Performance and
one Rx Application Service
ring and
Rx Board
Support,” Proc. of ACM
multiple RxRings SIGCOMM PRESTO 2008
Tx rings Rx Filtering Rx Filtering
Workshop, Seattle, WA, USA,
for each pp. 27-32.
forwardin
g core ATNAC 2010 28
Dynamic Adaptation
SW Routers & ACPI
[MHz]
Source: R. Bolla, R. Bruschi, A. Ranieri, “Green Support for PC-based Software Router:
Performance Evaluation and Modeling”, Proc. IEEE ICC 2009, Dresden, Germany, June 2009.
Best Paper Award.
ATNAC 2010 29
Sleeping / standby
Sleeping/standby approaches are used to
smartly and selectively drive unused
network/device portions to low standby
modes, and to wake them up only if
necessary.
However,
– since today’s networks and related services
and applications are designed to be
continuously and always available,
– standby modes have to be explicitly
supported with special techniques able to
maintain the “network presence” of
sleeping nodes/components.
ATNAC 2010 30
Sleeping / standby
Proxying the Network
Presence
Scenario: networked hosts (PCs, consumer
electronics, etc.);
Problem: when an end-host enters standby mode,
it freezes all network services, and it is
not able to maintain its network presence;
Idea: introduce a Network Connection Proxy
(NCP), which is devoted to maintain the
network presence of sleeping hosts.
Source: M. Allman, K. Christensen, B. Nordman, V. Paxson,
“Enabling an Energy-Efficient Future Internet Through
I want to sleep Zzzzz… Selectively Connected End Systems,” Proc. ACM SIGCOMM
HotNets, Atlanta , GA, Nov. 2007.
Wakeup/sleep Continuous
messages and full
connectivity
Application-
Sleeping host specific NCP Internet
messages
ATNAC 2010 31
Sleeping / standby
Proxying the Network
Presence
Scenario: Core Networks
Idea: put links, interfaces and part of nodes
(e.g., line-cards) to sleep
Problem: Network stability, convergence times
at multiple levels (e.g., MPLS traffic
engineering + IP routing)
Energy Consumption
Standby
ATNAC 2010 34
The Potential Impact
2015 - 2020 network forecast : device density and energy requirements
(example based on Italian network)
power consumption number of devices overall consumption
Home 10
(Wh) 17,500,000 1,533
(GWh/year)
Access 1,280 27,344 307
Metro/transport 6,000 1,750 92
Core 10,000 175 15
Sources : 1) BroadBand Code of Conduct V.3 (EC-JRC) and “inertial” technology improvements to 2015-2020 (home and
access cons.)
2) Telecom Italia
Network loaddevices
statistics measurements
and topology and evaluations
data (power consumption of metro/core network and number of
)
Home / Access target
customers per DSLAM 640 standby efficiency 85%
average usage of a network access 30% performance scaling efficiency 50%
average traffic when a user is connected 10% network-wide control efficiency 20%
air cooling/power supply efficiency 15%
Metro / Transport / Core
Sources : BroadBand Code of Conduct V.3 (EC-JRC) and
redundancy degree for metro/transport devices 13% technology improvements to 2015-2020.
redundancy degree for core devices 100%
redundancy degree of metro/transport device 100%
redundancy
links degree of core device links 50% Device internal sources of energy
average traffic load in metro networks 40%
average traffic load in core networks 40% consumption
Data Plane Control Plane Cooling/Power Supply
Home 79% 3% 18%
Access 84% 3% 13%
Source : forecast based on: carrier grade topologies;
traffic analysis and indicators (ETSI TR 102530, Metro/transport 73% 13% 14%
ODYSSEE) and projected traffic load. Core 54% 11% 35%
Sources : Information from vendors.
ATNAC 2010 35
The Potential Impact
5 - 2020 network forecast : device density and energy requirements
power number of overall consumption
Home 10
consumption 17,500,000
devices 1,533
(GWh/year)
Access 1,280
(Wh) 27,344 307
Metro/transport 6,000 1,750 92
Core 10,000 175 15
X Pstandby
ECONET target
standby efficiency 0.85 1 -
performance scaling efficiency 0.50 1 - Full Load Idle Standby
network-wide control efficiency 0.20 This value of standby power consumption refers to the case where
air cooling/power supply efficiency 0.15 1 - all cable-connected users are not active.
In the reality the ECONET technologies will enable access devices’
ports to selectively enter standby modes.
Device internal sources of energy
consumption
Data Control Cooling & Power
home Plane
95% Plane
3% Supply
2%
access 80.0% 3% 17%
metro/transport 73% 13% 14%
core 54% 11% 35%
Working at 100%
~
P= ϕ ϕ
~
ac tive activ e +
P standbyPstandby
Full load power Number of devices Overall full consumption Gains Energy gains with ECONET
Access 1consumption
,280 (W) 27 ,344
( relative ) 307
( GWh / year ) 70 % 213
technologies ( GWh / year )
The Potential Impact
Energy effic . target
Full Idl Standby Full Idl Standby Full Idl Standby Full Idl Standby
Load e Load e Load e Load e
ATNAC 2010 39
Ongoing Projects
ATNAC 2010 41
Ongoing Projects
ATNAC 2010 42
Ongoing Projects
ECONET Testbed
@ TELIT Test
Plant
ATNAC 2010 43
Ongoing Projects
TREND
Goals: integrating the activitiesof major
European players to quantitatively asses the
energy demand of current and future telecom
infrastructures, and to design energy-efficient,
scalable and sustainable future networks,
considering both the backbone, and the wireless
and wired access segments
Ø collecting data to assess the power consumption of
terminals, devices and infrastructures, to quantify
the current energy needs and their expected
evolution and scaling laws
Ø identifying energy-friendly technologies, devices,
protocols and architectures, and evaluating how
they can be introduced in operational networks
Ø defining new energy-aware network design criteria,
balancing the attention to optimal performance
and resource allocation with the minimization of
energy requirements
Ø experimentally proving
ATNAC the
2010effectiveness of the 44
proposed approaches
Ongoing Projects
TREND
ATNAC 2010 45
A few words about
research challenges …
Modelling green devices and networks
Ø For two purposes
• Control and optimization (fluid models,
performance bounds, closed-loop
strategies [but which timing?],
parametric optimization)
• Performance analysis (queueing models
with vacations and setup times at the
device level; traffic engineering [at
which layer?] at the network level),
indications on scalability
Experimental activity on test sites
Ø Real equipment
ATNAC 2010
Ø Network emulation 46
(e.g., router tester)
References
R. Bolla, R. Bruschi, F. Davoli, F. Cucchietti, “Energy Efficiency in the Future Internet:
A Survey of Existing Approaches and Trends in Energy-Aware Fixed Network
Infrastructures,“ IEEE Communications Surveys and Tutorials (IEEE COMST), to appear.
R. Bolla, R. Bruschi, K. Christensen, F. Cucchietti, F. Davoli, and S. Singh, “The
Potential Impact of Green Technologies in Next-Generation Wireline Networks – Is
There Room for Energy Saving Optimization?,” IEEE Communication Magazine (IEEE
COMMAG), Special Topic in “Green Communications ,” Nov. 2010, in press.
R. Bolla, R. Bruschi, A. Carrega, F. Davoli, “An Analytical Model for Designing and
Controlling New-Generation Green Devices,” Proc. 3rd IEEE Workshop on Green
Communications (GreenCom), co-located with GLOBECOM 2010, Miami, FL, USA. Winner of
the Best Paper Award.
R. Bolla, R. Bruschi, A. Cianfrani, M. Listanti, “Introducing Standby Capabilities into
Next-generation Network Devices,” Proc. ACM SIGCOMM Workshop on Programmable
Routers for Extensible Services of Tomorrow (ACM SIGCOMM PRESTO), Philadelphia , PA,
USA, Nov. 2010.
R. Bolla, R. Bruschi, A. Carrega, F. Davoli, "Theoretical and technological limitations
of power scaling in network devices", Proc. Australasian Telecommunication Networks
and Applications Conf. 2010 (ATNAC 2010), Auckland , New Zealand, Nov. 2010.
R. Bolla, R. Bruschi, A. Ranieri “Green Support for PC-based Software Router: Performance
Evaluation and Modeling,” Proc. 2009 IEEE International Conference on
Communications (IEEE ICC 2009), Dresden, Germany, June 2009. Winner of the Best
Paper Award.
ATNAC 2010 47
References
R. Bolla, R. Bruschi and F. Davoli, “Energy-aware Performance
Optimization for Next Generation Green Network Equipment”,
Proc. ACM SIGCOMM 2009 Workshop on Programmable Routers
for Extensible Services of Tomorrow (ACM PRESTO 2009),
Barcelona, Spain, Aug. 2009, pp. 49-54.
R. Bolla, R. Bruschi, F. Davoli, “Energy-aware Resource
Adaptation for Next Generation Network Equipment,” Proc.
2009 IEEE Global Communications Conference (IEEE GLOBECOM
2009), Honolulu, Hawaii, USA, Dec. 2009.
R. Bolla, R. Bruschi, A. Carrega, F. Davoli, “Green Network
Technologies and the Art of Trading-off,” submitted to IEEE
Infocom 2011.
R. Bolla, R. Bruschi, A. Cianfrani, M. Listanti, “Backbone
Networks to Sleep,” submitted to IEEE Network Magazine,
Special Issue on “Green Networking”.
ATNAC 2010 48