The All Optical Network

Why we need it how and when we get it David Payne BT

Co-inventor of TPON and Principal Consultant, Optical Networks

Thanks to the optical team:

Alan Hill Alan McGuire Albert Rafel Andrew Lord Dave McCartney Derek Nesset Ed Sikora Ian hope Ivan Boyd John Wright Justin Kang Kristan Farrow
David Payne BT 2007

Martin Wade Paul Tomlinson Paul Wright Peter Chidgey Peter Healey Phil Barker Ruben Gorena Russell Davey Shamil Appathurai Steve Hornung Tim Gilfedder Yu-rong Zhou

Why optical networks? the physics perspective

The electro-magnetic spectrum

The optical fibre window 1.87x1014 to 2.38x1014 Higher carrier frequencies (1600nm to 1260nm)

David Payne BT 2007

Why optical networks? the physics perspective

Channel Capacity Limits
10 24 1.0E+20

Relative Channel Capacity

10 22 1.0E+18 10 20 1.0E+16 10 18 1.0E+14 10 16 1.0E+12 10 14 1.0E+10 10 12 1.0E+08 10 10 1.0E+06

1.0E+04 10 8 1.0E+02 10 6
1.00E+05 1.00E+06

Capacity per Hz Channel capacity bits/s Fibre spectrum Cumulative capacity

The optical fibre window 50 THz ~1000Tb/s

1.00E+07

1.00E+08

1.00E+09

1.00E+10

1.00E+11

1.00E+12

1.00E+13

1.00E+14

1.00E+15

1.00E+16

1.00E+17

1.00E+18

1.00E+19

1.0E-02 10 2 1.0E-04 1

Frequency (Hz)

David Payne BT 2007

1.00E+20

1.0E+00 10 4

Why optical networks? the physics perspective Summary 1

It will not be worth going beyond ~3nm wavelength (~1017Hz) due to fundamental quantum noise limitations. Current generation optical communications systems are within two orders of magnitude of that limit in information capacity terms. But we have no technology currently being developed that exceeds the capability of optical fibre or that could exploit the higher frequencies. In the past we have had a fundamentally better technology (usually more expensive) for core networks than the access network. With fibre ubiquitously deployed in access, metro and core networks this is no longer the case.

Therefore end users of a ubiquitous fibre network can never have the full bandwidth of a fibre dedicated to them!
This has consequences for the choice of network architecture.

David Payne BT 2007

Why optical networks? the commercial perspective

Drivers
Meeting customer demands for new high bandwidth services New revenue generation Reducing operational costs Meeting competition threats Staying locally or internationally competitive
Attract inward investment

David Payne BT 2007

Why optical networks? the commercial perspective

Bandwidth demand
New services including:
video, hdtv etc. Large file transfer eg. hi-res image & video content, photographs etc. Mass market network storage services Distributed servers Thin client computing Etc.

End users need access to high speed pipes

To get low delay, fast file transfer & rapid response time But they dont need the bandwidth all of the time. Therefore redistribute the unused bandwidth DBA for all services!

David Payne BT 2007

Impatience Index
100% 90% 80% 70% % Customers that wait 60% 50% 40% 30% 20% 10% 0% 0 5 10 15 20 25 30 35
00 20 a irc (c Up ) -3 al Di 02 20 a irc (c nd ba B rly 6) Ea 200 nd ( Bba

time to download (Secs)

David Payne BT 2007

Time to transmit files

100 days 10 days 1 day 8 hours Time to transmit 1 hour 10 min 1min 10 sec 1 sec 200ms 1 10 100 1,000 10,000 Postal Service

Physical transport faster!

b/s 56k m ort ode /s st p M cu Mb b/s 0.5 b/s orts /s 0M MB st p Up n 2M 0 L cu )3 P1 D S dow Up b/s TT A F 1G SL SL TP (VD AD FT TC FT

} }
}

Delay a major barrier

Customers get impatient Customers accept delay Instantaneous

CD

DVD HDTV

100,000

File size (Mbytes)

David Payne BT 2007

Example Service Scenario

Streaming Services Internet Surfing IP HDTV HR IP HDTV HR IP HDTV LR IP HDTV LR IP SDTV IP SDTV IP SDTV IP SDTV Video comms Camcorder Web Cam comms Voice Voice Voice Thin client computing Thin client computing File Transfers e-mail e-mail Photos Photos Video clips Video clips Music Music Documents Documents Software Software Web site uploads Web site uploads Peer to peer file sharing Peer to peer file sharing Storage services Storage services 0% 0% 30% 20% 20% 20% 30% 20% 10% 5% 5% 2% 3% 0% 5% 3% 5% 3% 80% 60% 70% 30% 60% 40% 40% 30% 20% 15% 10% 5% 10% 1% 40% 20% 30% 15% 10% 10% 5% 5% 15% 15% 5% 5% 5% 5% 3% 2% 1% 1% 40% 30% 1% 1% 20% 15% 10% 10% 25% 25% 15% 15% 10% 10% 8% 5% 5% 5% 50% 40% 5% 2% Probability of takeup 2008 60% 0% 0% 10% 10% 90% 80% 40% 20% 2% 10% 100% 75% 50% 0% 0% 2017 100% 90% 80% 30% 12% 50% 25% 20% 10% 15% 15% 100% 75% 50% 30% 20% Probability of using service in busy hour 2008 40% 70% 50% 30% 10% 70% 50% 30% 10% 20% 20% 20% 15% 10% 10% 10% 2017 70% 80% 60% 40% 20% 80% 60% 40% 20% 30% 30% 20% 15% 10% 50% 50%

Table 1 Example service scenario 1

David Payne BT 2007

Bandwidth Usage Results

From example service scenario 2008
This chart shows the distribution of the average FTTP user bandwidth for the service scenario parameters for year 2008. The mean is 2.0 Mb/s Note the significant probability that some users do not use any services in the busy hour.

This chart shows the distribution of the maximum user bandwidth if the user had simultaneously used all the services they had signed up for. This is a dumb way of defining user bandwidth and would push the mean to 29.9 Mb/s 15 times the required level if statistical multiplexing is taken into account.

David Payne BT 2007

Bandwidth Usage Results

From example service scenario 2017
This chart shows the distribution of the average user bandwidth for the service scenario parameters for year 2017. The mean traffic has grown from 2.0 Mb/s 7.5 Mb/s. The probability of users using no services has reduced significantly but is still large enough to remain the mode of the distribution.

This chart shows the distribution of the maximum user bandwidth if the user had simultaneously used all the services they had signed up for. This is the dumb way of defining user bandwidth and would push the mean to 82 Mb/s, 11 times the required level if statistical multiplexing is taken into account.

David Payne BT 2007

End user bandwidth growth

Effective end user average bandwidth
100.0 90.0 80.0 70.0 Mb/s Mb/s 60.0 50.0 40.0 30.0 20.0 10.0 0.0 0 2 4 Years 6 8 10
Peak user bandwidth Average peak user bandwidth Average user bandwidth

David Payne BT 2007

End user access pipes

Streaming mode
Access network bandwidth Pipe (eg. GPON DS bandwidth) Spare bandwidth pool = access pipe size used bandwidth

End user 1

End user 2 Spare bandwidth not used in streaming mode

End user 3

user 1 user 5 user 4

CPs

End user 4

user 3 user 2 user 1

End user 5 Streaming mode Used bandwidth profile

David Payne BT 2007

End user access pipes

Burst-mode
Access network bandwidth Pipe (eg. GPON DS bandwidth)

End user 1

End user 2

Burst-mode Used bandwidth profile All spare bandwidth is used to deliver files as fast as possible

End user 3

CPs

End user 4

End user 5 Spare bandwidth shared between active users Streaming mode Used bandwidth profile

David Payne BT 2007

Improvements in End User experience from burst-mode operation

512 Customers, 10G PON Bandwidth, 1000M Access Port, No Broadcast
700,000,000.00 Average Session Bandwidth (bits/s) Average Session bandwidth Mb/s

700

Streamed Scenario

Burst Mode Scenario

600,000,000.00 600

500,000,000.00 500

400,000,000.00 400

300,000,000.00 300

200,000,000.00 200

10% 15% 20% 30% 40% 50% 60% 70% 80% 90%

100,000,000.00 100

0 0.00
ne e ne e eo eo o o s s es on on Vi de ice Vi de ic e ph o ph o Vi d op h op h Vi d ic rv Se rv rv Te le Te le HD SD Vi de Vi de SD H c c Se rv em iu m Pr Se Se D ice s

iu m

Ba si

Pr

David Payne BT 2007

em

Ba si

Why optical networks? - commercial & services perspective

Summary 2
There are a number of powerful drivers for all optical networking particularly Fibre to the Premises (FTTP) networks: Attracting inward investment Operational cost savings New services for end users New revenues into the industry Improved end user experience click and its there! Users need high speed access + higher ave. bandwidth
David Payne BT 2007

So why arent optical networks everywhere?

Barriers
High upfront capital investment
Civil infrastructure build costs Also time taken to build

Increased opex costs if legacy not replaced Uncertain take-up rates Uncertain revenues from new broadband services Financial barriers

David Payne BT 2007

Todays optical access network options

Point to Point Solutions
Point to point fibre WDM PON Active Star

Passive Optical Network:

BPON GEPON GPON
David Payne BT 2007

Options for optical access

Point to point fibre.
Non-FTTP Customers

Big Business Customer

Local Exchange

backhaul/metro network

Cabinet
Cabinet

FTTP Customers

David Payne BT 2007

Options for optical access

Point to point fibre active cabinet.
Non-FTTP Customers

Big Business Customer

Local Exchange

backhaul/metro network
Cabinet

Active star
Cabinet

Active star

FTTP Customers

David Payne BT 2007

Options for optical access

WDM PON.
Non-FTTP Customers

Big Business Customer

WDM Splitter

Local Exchange

Cabinet

backhaul/metro network
WDM Multiplexer
Street MSAN

copper

FTTP Customers

David Payne BT 2007

Pt-Pt optical access network options

Point to Point Solutions:
Point to point fibre to exchange + High bandwidth per customer + Good security - Large fibre count cables - two opto-electronic modules per customer - Opex and capex issues Active Star + Shares feeder fibre - Active street electronics opex issues WDM PON Logical point to point + WDM enables some fibre sharing - Requires advanced WDM optics - Nails wavelengths to customers very inefficient use of optical spectrum

No bandwidth sharing Separates access and metro networks Keeps traditional architectures and cost structure Keeps electronic nodes - either in exchanges, in street nodes or both

David Payne BT 2007

Options for optical access

Passive Optical Network (PON)
Non-FTTP Customers

Big Business Customer

Local Exchange

Cabinet

/s 2.5Gb 5Gb/s 1.25/2.

backhaul/metro network

~32 way split

Cabinet MSAN

copper
FTTP Customers

David Payne BT 2007

PON options
Passive Optical Networks:
BPON GEPON GPON A-GPON LR-PON Features + Infrastructure sharing + Passive infrastructure no active street electronics + Enables dynamic bandwidth sharing + Lower cost - particularly extended/long reach systems + PONs can be protected economically

David Payne BT 2007

Options for optical access

Summary 3: Claims and Pros & Cons of PON v Pt-Pt
Common Claim: Pt to Pt is more future proof than PON
Customers must share fibre bandwidth by multiplexing bandwidth from access fibres into core fibres. Therefore customer bandwidth is always less than the fibre bandwidth! All the fibre bandwidth can never be allocated to individual customers No fundamental difference between the End User bandwidth upgrade potential for Pt to Pt or PON architectures PON is a more flexible architecture - can be considered to be the most future proof architecture. PONs can support legacy protocols (TDM, ATM etc.) if required on a common transmission protocol. PONs can support multi-casting/broadcasting natively (only one copy of a channel require for N customers).
Customers can access many tens of channels simultaneously without increasing access and backhaul bandwidth requirements.

WDM can be used to up-grade PON capacity and user bandwidth to the limits of the fibre capacity.

The question becomes:

What is the simplest, lowest cost and lowest power consumption multiplexing technique?

Pt-Pt uses an electronic multiplexer with a port per customer on the access side. The equipment requires power and is housed in a building or street cabinet PON systems use passive optical multiplexing which consumes no power and can fit into external plant.

In practice both Point to Point and Passive Optical Networks will co-exist in the network for different market situations. However PON will be the mass market solution.
David Payne BT 2007

Economic Issues
FTTP drives a bandwidth demand paradigm shift New broadband services in particular HDTV/IPTV enabled by FTTP will massively increase demand on the metro/core network DSL has been successful at enabling good user experience for many non-video Internet applications But it will struggle to provide a future basket of services (e.g. HDTV, 3D applications, real time video, peer to peer file sharing including video etc.) So currently were at the onset of a worldwide access fibre program What would be the impact of an access bandwidth explosion driven by FTTP on the end to end network?
David Payne BT 2007

Bandwidth Growth The Margin Challenge

100 90

Relative growth

80 70 60 50 40 30 20 10 0 2000 1 2001 2 2002 3 2003 4 2004 5 2005 6 2006 7

Bandwidth Bandwidth

Costs

Greater bandwidths - New services - Maintain/grow revenues But costs rise faster

100 90 80

Relative growth

70 60 50 40 30 20 10 0 2000 1 2001 2 2002 3 2003 4 2004 5 2005 6 2006 7

Revenues

Revenues

Margins are eroded

Costs

Incremental Costs

years
David Payne BT 2007

The economics of current FTTP solutions do not look attractive for mass market deployment. Cost of bandwidth growth enabled by FTTP can easily exceed revenue growth. Can new network architectures enable FTTP and future bandwidth growth to become economically viable? Must consider end to end network solutions: Access, Metro & Core together.

David Payne BT 2007

BTs 21st Century Network

Inner Core PoPs

Outer Core PoPs

Electronic conversion when crossing layer

Metro PoPs

Access PoPs with WDM

Access PoPs

David Payne BT 2007

Will this architecture scale economically with FTTP?

Doubling the bandwidth more than doubles the electronics
Electronics cost reductions are not fast enough Bandwidth cant become cheap enough whilst maintaining profit No long-term incentive to provide high bandwidth services

Too many exchange buildings and to much equipment

CapEx and OpEx too high Environmentally unfriendly with large power usage Network optimised around existing / legacy / known traffic / services

Not configured for content distribution or highly dynamic bandwidth

The architecture has to change The architecture has to change

David Payne BT 2007

Integrated access and backhaul network

Long reach access
Non-FTTP Customers

100km

Big Business Customer

EDFA technology constrains operation to C band wavelength range

backhaul/metro Cabinet
/s 10Gb b/s 2.5/10G

~500 to 1000 way split

Cabinet MSAN

FTTP Customers

copper

All ONUs have fixed band-pass blocking filters which only select the initial operating wavelength and block all others. When WDM is added at a later stage the new ONUs will have a band pass filter at one of the additional wavelengths

David Payne BT 2007

Integrated access and backhaul add circuit switched

optical core
Long reach access
Non-FTTP Customers

Intelligent Photonic Inner core Network

100km

Big Business Customer

EDFA technology constrains operation to C band wavelength range

backhaul/metro
Cabinet

Metro nodes

/s 10Gb b/s 2.5/10G

~1000 way split

Cabinet MSAN All ONUs have fixed band-pass blocking filters which only select the initial operating wavelength and block all others. When WDM is added at a later stage the new ONUs will have a band pass filter at one of the additional wavelengths

FTTP Customers

copper

Optical switches

David Payne BT 2007

Integrated access and backhaul with photonic core

Longer Term Vision Hybrid WDM/TDM + Flexible wavelength assignment Intelligent
Long reach access
Photonic Inner core Network

Big Business Customer

/s 10Gb 10Gb/s

New amplifier technology (e.g. quantum dot) enables operation across all the fibre spectrum

backhaul/metro

Metro nodes

Cabinet

Cabinet MSAN

Power splitter (not WDM splitter) to enable any or combination of s to any customer

FTTP Customers

copper

Optical switches

Network Reduced to ~100 exchanges

Tunable & self install ONU (for residential customers)

David Payne BT 2007

Optics-centric architecture
All nodes both access and core

~100 node alloptical core Long reach WDM/TDM PON access

Inner / outer core hierarchy scrapped

No backhaul / metro network

Access nodes replaced by passive optics + optical amplifiers

David Payne BT 2007

Evolution scenario of BT network

David Payne BT 2007

Operational & Environmental Benefits

Based on Ipswich Exchange serving ~15,000 customers Today 21C MSAN 900 racks 20 racks Long Reach PON <1 rack

~800 kw 50-100 kw

100W

David Payne BT 2007

1 per ~1000 customers

Amplified GPON
Adding amplifiers to GPON can be an interim solution for LR-PON

ONU

Local Exchange
ONU

Tx OLT Rx

ONU 4x32-way Split

4 X 4
Tx Rx

OLT S

ONU

128-way total split

60km (100km)

David Payne BT 2007

Status of Amplified GPON

Extender box standardisation active topic in FSAN Standard expected early 2008. First development prototype of commercial amplified GPON system (Motorola) now in BTs Labs. Blocking filter standard for WDM upgrade standard agreed. Could see standards compliant systems commercially available inside two years.

David Payne BT 2007

Status of NGA Long Reach PON (LR-PON)

Principles demonstrated 4 years ago by BT Active topic in NGA working group in FSAN System components actively being developed in collaborative projects. First demonstration prototype from major supplier (Nokia-Siemens) in BT Labs September 2007. Evolution path from GPON to LR-PON determined.

David Payne BT 2007

Nokia-Siemens LR-PON Concept

512 subscribers 512 subscribers

512 subscribers

100km
10Gbit/s
Access Access Exchange Exchange
Access Exchange Access Exchange

Node
OLT

2.5Gbit/s (10G ready)

up to 2048 end users sharing backhaul fibre

512 subscribers

David Payne BT 2007

Economic comparisons
Cash Flow
2.00

1.50
Option 3: Std GPON + 21cn GE B'haul

1.00 B illio n s

Option 0: Pt-Pt fibre GE B'haul

Option 3: Std GPON + 21cn GE B'haul

0.50

Option 4: Amplified GPON Option 5: LR-PON

0.00 0 1 2 3 4 5 6 7 8 9

Option 8: VDSL Cab + 21cn GE B'haul

10

11

12

-0.50

-1.00 Years

David Payne BT 2007

Comparison of time to breakeven v sustained end user bandwidth

Sustained Cust B'wdth v Time to Breakeven
16.00

14.00

12.00

Time to +ve cash flow (years)

10.00

8.00

6.00

VDSLCab upstream bandwidth exceeded

VDSLCab downstream bandwidth exceeded

Std GPON A-GPON LR-PON VDSLCab+21cn

4.00

2.00

0.00 0.1 1 B'width Mb/s 10 100

Assumes no correlation of revenue with bandwidth + self install ONT

David Payne BT 2007

Time to breakeven v PON split size

Time to breakeven v split
8.00

7.00

6.00

Time to breakeven

5.00 LR-PON - 4.00 A-GPON Std GPON 3.00

2.00

1.00

0.00 10 100 PON split size 1000

End user sustained bandwidth ~10 Mb/s

David Payne BT 2007

Optical access & metro network Summary 4

Todays optical access and separate metro/backhaul network solutions not economic for mass deployment.
Cost of bandwidth growth would exceed any revenue growth.

Need to reduce amount of electronic equipment and number of network nodes. Achieved by combining access and metro/backhaul. into one network
Long Reach PON proposed as possible solution. Amplified GPON as interim solution. LR-PON economics look promising, FTTP can be competitive with FTTCab particularly as end user bandwidths increase.

David Payne BT 2007

Impact of FTTP & long-reach access on core

Some orders of magnitude: ~10 million customers, 100km long-reach access, 10 Mb/s sustained rate busy hour ~100 core nodes Total sustained traffic = 100 Tb/s Traffic per node = 1 Tb/s
If equally distributed amongst core nodes

Average traffic between each set of node pairs = 10 Gb/s

Assuming little turn-around traffic, no protection, no bandwidth contention, etc.

Assign 1 between all node pairs

No need for grooming / intermediate electronics.

David Payne BT 2007

Long-reach access core Metro/core-node structure

Ave 100,000 customers
500 per PON

To other nodes

Metro/core node
OLT OLT OLT OLT OLT OLT

Average 100 lambdas

Average 200 OLTs

Aggregation / routing

10km

WDM backhaul

OLT OLT OLT OLT OLT OLT OLT OLT OLT OLT OLT

Duct routes out of exchange 100km

David Payne BT 2007

Optical switching

Effect on Locations for Head-Ends (OLTs)

UK With All Exchanges

UK with~100 21cn nodes

David Payne BT 2007

Numbers of nodes bypassed

Traffic Paths
14% 12% 10% 8% 6% 4% 2% 0% 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Physical Hops

Assumes 1 between all node pairs Significant amount of node bypass Therefore need cheap large scale node bypass optics to make this architecture cost in

David Payne BT 2007

Potential future UK all-optical core network

All-optical meshed core Light-paths set up between all core nodes Optical path bypasses intermediate nodes using no electronics

David Payne BT 2007

Flat versus hierarchical core

Full logical mesh between all routers to create a flat core
Assumes routers/switches can handle connectivity to ~100 nodes

Optical switching to allow for flexibility, meshed restoration, etc.

Optical interfaces much cheaper than router interfaces Router/switches smaller do not need to handle through traffic

Hierarchical core allows for intermediate electronic grooming for limited traffic levels

Router OXC?

Router

Router OXC

Router

outer core

OXC?

flat core

OXC

Router OXC?

Router OXC?

Router OXC?

Router OXC

Router OXC

Router OXC

inner core

David Payne BT 2007

Flat vs hierarchical core

Network Hierarchy Comparsion
60000 50000 40000 Line Cards 30000 20000 10000 0 0 20 40 60 80 100 120 140 160 180 200 Network Traffic (Tb/s) Two-Tier Flat

2 tier model assumes 20 inner core nodes Flat architecture preferred when wavelength fill > 40%

David Payne BT 2007

Network node dimensioning

Nodal Capacity
3500 3000 2500 2000 1500 1000 500 0 Nodes Total Wavelength Ports Utilised Wavelength Ports

Assumption of long reach access + flat all-optical core leads to huge core transport + very large optical switches Optical switch capacity up to 40 fibre pairs, assuming 80 channel DWDM Up to 20 fibre pairs on busy links

David Payne BT 2007

Possible new optical core node architecture

Scaleable to many fibre pairs Fibre switch enables fibre bypass of node. eg. NxN MEMs

To other nodes Fibre switch

To other nodes Some fibres diverted to wavelength switch for local node wavelength add/drop and also possible wavelength grooming Wavelength switch could use WSS technology

Wavelength switch To access side of node Electronic (IP / Ethernet)

Cascading nodes leads to requirement for ultra low loss

David Payne BT 2007

How do we provide such huge optical bandwidths?

How do we optimise split between Electrical switching, Wavelength Division Multiplexing (WDM) and Fibre Division Multiplexing (FiDM)? Low fibre count solution - 40Gb/s, 100Gb/s, C-L-Sbands, high spectral efficiency
Eggs in one basket

Multiple fibres Cheaper simpler 10G transmission

More lambdas more optical switching flexibility Multiple line systems = more line equipment More resilient?

Many optical technologies presented at conferences OFC, ECOC etc over last 10 years now need to be considered. Economics and resilience key to deciding best approach
David Payne BT 2007

Optical Core Summary 5

Fibre access enables unprecedented bandwidth growth.
Challenging economics needs new architectures the bandwidth growth will also pressurise the core

All-optical flat architecture with no separate metro network provides economic scalability. In the UK this equates to:
~ 100 core nodes with over 10Tb/s on some links Up to 1500 10G wavelengths bypassing nodes

Large optical switches at fibre and wavelength level will be needed High bit rates / multiple wavebands (C,L,S) will all be required Economics and resilience at the fibre, cable and duct level will determine optimal design

David Payne BT 2007

Technology and evolution timeline

Backhaul GPON WDM SDH Ethernet

amplified GPON (60 km) Greenfield access GPON Powered Cabinets

+ scale protocol to 1024 split +10Gbit/s

10 Gbit/s LR-PON
(100+ km)

+colourless ONUs +WDM in backhaul

WDM LR-PON

+tunable optics

Flexible LR-PON

NonGreenfield access

PIEMAN

2008

2012

David Payne BT 2007

Overall Summary
We have the opportunity with the all optical network of radically changing the architecture to:
Greatly reduce the capital expenditure required for a true broadband network Massively reduce operational costs. Produce an environmentally friendly network >90% reductions in energy consumption. Revolutionise the customer experience. Click and its there! Provide a reusable and continuously upgradeable physical infrastructure for the Next Generation 21st Century Network with true broadband capability!

David Payne BT 2007

Thank you

David Payne BT 2007