1. Greening your Edge
Are you a power savior or a power pig?
Francois Lemarchand <francois.lemarchand@ericsson.com>

2. An Initial take:
Per node power consumption
14000
12000
10000
8000
6000
4000
2000
0
DSLAM Aggregation Edge Core
Most savings appear at first to be done on the Core Nodes
© Ericsson AB 2009 | Ericsson Internal | X (X) | Date

3. Looking at efficiency:
Power versus bandwidth
120 * A given flow will need to cross multiple of these nodes in the network
Driven by DSL
100 loop modulation
80 Driven by high touch
service processing
60
40
20
0
DSLAM *
Aggregation Edge Core*
But the biggest potential probably lies into the access layer
© Ericsson AB 2009 | Ericsson Internal | X (X) | Date

4. Silicon process evolution only
compensate bandwidth growth
› Each generation of silicon process brings twice the
amount of transistor logic in the same cost / footprint /
power.
– This could mean halving the power consumption every 2 years
– But the Internet and data bandwidth consumption is following
the same pattern (rough doubling every two years)
– Market pressure has caused to reinvest this capital in network
performance instead of power savings
Technology will bring significant bandwidth efficiency
© Ericsson AB 2009 | Ericsson Internal | X (X) | Date

5. Power savvy technology design
is need for a visible change
› The key is to dynamically put to idle
unused capacity
– CPU manufacturer have shown the potential of this
option. It will come to Network Processor over time:
› Semi dynamic: i.e. putting down to idle backup
linecards / or silicon handling specific port group
› Fully dynamic: Clock / power adjustment based on
dynamic traffic load
› At the network level critical savings are
must come from the access
– High promises from adaptive DSL modulation
Node level power savings are dependant from innovations
© Ericsson AB 2009 | Ericsson Internal | X (X) | Date

6. Network architecture
optimization
› Optimization of the power consumption of
individual nodes lead to significant Opex
savings
– I.e. Loaded cost of KW/H per year up to $2000
– 5KW node over 5 years = $50K
› Optimization of the network architecture
can lead to additional Opex but also
Capex savings
› While node level optimization is
dependant on the vendors roadmap the
operators are in full control of the network
architecture

7. Cost/POWER per bit hierarchy
Optimize network complexity
Mobile
Voice
Video
Internet
Enterprise VPN
Enterprise ELINE, ELAN
Legacy ATM / TDM transport
› Service Node – L3/L4/L5
service point: Fixed (BNG) or
PGW SBG CDN DPI IPS
Mobile packet GW, Enterprise
L3PE w/Security, Video Edge
IP Services (Subscribers,
IP Flows, Application) w/Caching etc…
› L3 Edge / Border Node: L2
termination and IP transport
IP Routing, L3VPN
› L2 Edge / BN: L2 service
point: PWE ingress/egress,
PWE / CES / Bridging
L2 interworking, QoS, security
MPLS › Core / Transport Node: pure
MPLS L2 switching, no
edge services.
› Optical L1 moving to
Optical
© Ericsson AB 2010 | Ericsson Internal | Aggregation Routing Strategy (draft12) | 17 feb 2010
OOO ROADM

8. Core to access scale factor
Increased returns in the outer ring
1-10,000’s of Aggregation nodes
10-100,000’s of Access nodes
100-1000’s of Service Edge nodes
10-100’s of Core routers & optical switches
Millions of connections, devices
© Ericsson AB 2010 | Ericsson Internal | Aggregation Routing Strategy (draft12) | 17 feb 2010

9. Fully Consolidated Edge Model
another angle at power savings
Users & Access IP Edge Services
Devices Mobility
Functions
DPI
Functions Internet & VOD
BRAS
Functions
Video/IPTV
SBC
Functions VoIP
Eth Agg
Functions
L2/L3
VPN
L3 PE
Functions
Servers
L2 PE
Functions Mobile
Services
Need to balance consolidation with induced operational impacts

10. Intelligence placement
balance scalability and power cost
+ Capex - Peering opt. intelligence
+ Opex - Content opt.
BW per node
Sweet
Spot
# of Nodes
› Over time centralized intelligent functions had been distributed further toward the
edge in order to scale with the bandwidth constraints. But it needs to be
balanced by the induced opex and capex and power efficiency cost.
› Technology improvements do not change fundamentally that balance
– Allow to build smarter functions with limited cost impact on the access nodes
– But also allows to build bigger Service Nodes for the same price
– Besides BW intelligence level also keeps increasing (LI, DPI, Mobility, v6, NAT…)
Keeping the aggregation/access simple optimizes the power
© Ericsson AB 2010 | Ericsson Internal | Aggregation Routing Strategy (draft12) | 17 feb 2010

11. Transport & Metro convergence
Packet optical transport (POTP)
Fixed GW
Native packet mux & mcast
L2PE L2PE
PGW/RNC
Core Core
T3CO L2PE L2PE
Cell site T2CO Enter. L2L3
T2CO T1CO S PoP
T3CO Core
Core
T3CO T2CO T1CO S PoP
T3CO T2CO IP/MPLS routing
Packet & Optical Transport Platform
All Optical transport
10,000’s+ T3CO’s 10,000’s T2CO’s 1000’s T1 CO’s 100’s Service PoPs 10’s Core PoPs Few National PoPs
› Provides the BW efficiency of packet based multiplexing in the
transport layer for native packet services or emulated TDM circuits
› Allows to subsume overlay metro network capabilities into the transport
layer – a single layer of transport equipment to carry enterprise,
residential, fixed, mobile and wholesale traffic.
› Running a single integrated control plane / NMS across the packet and
optical layer allows to optimize the mapping between the optical
resources and the packet network transit
© Ericsson AB 2010 | Ericsson Internal | Aggregation Routing Strategy (draft12) | 17 feb 2010

12. Optimize packet with optical
But avoids the overlay Model
L3PE L3PE L3PE
OTN
switch
MPLS
Internet
Peering
SDH/OTN
Optical
› OTN layers allows to efficiently multiplex packet and TDM transparent
services into the same lambda.
› As packet services are becoming predominant it makes sense to map
them directly over a lambda – integrating the WDM/OTN optics directly
into the packet switching function to optimize the processing
› Further optimizations are possible by doing a selective bypass of
certain traffic flows (i.e. toward centralized video hosting or internet
peering points)
© Ericsson AB 2010 | Ericsson Internal | Aggregation Routing Strategy (draft12) | 17 feb 2010

13. There is an elephant in the room
capex is in but the opex is out?
DSL Routing Firewall VOIP/SIP WIFI
= 10-15W
Modem NAT
Home Gateway
Set Top Box MPEG decoder Hard Drive
= 15-20W
› Over time the home gateway has grown in functionality / complexity.
Today’s home GW can consume up to 15W. With the introduction of
IPTV decoders the power per home is rising to the 20-30W range.
› Today’s HGW & STB have received a limited focus on developing
power savings versus functionalities => always on
› It is urgent to introduce power saving functions
© Ericsson AB 2009 | Ericsson Internal | X (X) | Date

14. What does it mean when applied
to the 70M US Broadband lines?
2 Millions Households powered during a year
10 Millions Tons of Co2 emission
© Ericsson AB 2009 | Ericsson Internal | X (X) | Date

15. Power Per subscriber
Access drives 95% of the power
35 95% of fixed
30 broadband network
power consumption
25
20 5% power
15
Epsilon
10
5 1W 0.01W 0.0001W
0
Home DSLAM Edge Core
How do drive back household power requirements?
© Ericsson AB 2009 | Ericsson Internal | X (X) | Date

16. Simplification of the home 1/4
Network based VOIP GW function
POTS phone
Network SIP Client
HGW based SIP integration in the
client MSAN
Aggregation
Network
@
HGW
Access Node BNG Edge
STB
› A number of operators have chosen to integrate the VOIP client into the
Home GW. But centralization into a network equipement such as the
MSAN can provide significant opex & power savings
› It also facilitates the migration of the fixed voice customer to a cost
effective VOIP access GW.
© Ericsson AB 2009 | Ericsson Internal | X (X) | Date

17. Simplification of the home 1/4
Carrier Grade NAT and Bridge HGW
Configuration as
transparent bridge
Private IPv4 to Carrier Grade NAT
public IPv4 NAT At the BNG
Aggregation
Network
@
HGW
Access Node BNG Edge
STB
› In order to address IPv4 address depletion carriers will progressively
introduce a network based NAT function to allows sharing of NAT
public pools between more customers (native IPv6 in the longer term)
› This provides an opportunity to simplified the HGW and perform L3/L4
functions at the BNG Edge instead. HGW back to a bridge modem.
© Ericsson AB 2009 | Ericsson Internal | X (X) | Date

18. Simplification of the home 1/4
Network based PVR
Network PVR
Aggregation
Network
@
HGW
Personal PVR
Access Node BNG Edge
STB
› Network based PVR & Catch’up TV functions get an increased level of
popularity and acceptance by the content providers
› Incidentally it also provides significant capex, opex and power reduction
by removing the requirement to support hard drive and recording
functions on the STB.
© Ericsson AB 2009 | Ericsson Internal | X (X) | Date

19. Simplification of the home 1/4
Open IPTV & Native IPTV clients
Aggregation
Network
@
TV with integrated
IPTV capabilities HGW
Access Node BNG Edge
X STB
› There is a growing commitment of the industry – Service Providers,
Telecom Vendors & Consumer electronic companies to support
standard IPTV specifications and interfaces (Open IPTV Forum)
› One of the benefit is to allow the TV set providers to integrate native
IPTV functions into the TV set – and for operators to suppress STB
© Ericsson AB 2009 | Ericsson Internal | X (X) | Date

20. Green your access
it starts at the Edge
› Vendors will gradually introduce green
technology design over time
› Operators should consider their network
architecture design with an holistic
approach to bring power / capex and
opex savings.
› Priorities must be set to the home /
access / aggregation where volumes can
drive the most significant savings
› The IP Edge is as a power saving enabler
– key to simplify the architecture of the
home, access and aggregation layers.