ONE'06 Optical Network Europe

1
ONE'06 Optical Network Europe

• Flexible optical networks the end of dumb pipes?

ONE06 Cannes 2006
2
Agenda

• The story so far
• How long?
• How flexible?
• How integrated?
• How smart?
• Conclusions

3
Where we are now

• At one time, operators regarded their optical
  networks as plumbing - fat pipes to join
  switching nodes
• Then, as well as steadily increasing capacity,
  the optical equipment gained
• Network resilience, with SDH and OTN (optical
  transport network)
• Packet-friendly (Ethernet) features - with Next
  Generation SDH
• Ethernet and MPLS switching - with Ethernet
  Transport products
• Wavelength switching - with ROADMs (remotely
  reconfigurable optical add-drop multiplexers)
• And sometimes several of these in the same box

• Now operators have a richer range of choices
• Select the right feature mix, not a swiss army
  knife
• Network designers have to optimise the whole
  network, not just each layer by layer

Increased value lies in best practice network
design
4
Context - public network technology layers
Services
Layers 4-7
Video
Voice
Data
Layer Role
3 Routing
IP
2 Protocol-specific data switching
Frame Relay
ATM
RPR
Ethernet
Packets
5
Growth of deployed transport capability per
bearer - a 150 year trend will continue?
Equivalent bits per second
Fibre optics
1 Tbit/s

• Historical growth 30 per year

100 Mbit/s
10 kbit/s
10 bit/s
1850
1900
1950
2000
6
Agenda

• The story so far
• How long?
• How flexible?
• How integrated?
• How smart?
• Conclusions

7
WDM line systems - a range of solutions

• Max distances without regeneration at 2.5 / 10G
• CWDM systems - typically 40-100km
• DWDM systems - typically 80 - 3000km
• Metro DWDM - typically to 300km
• (Ultra) long haul - to around 3000km, advanced
  pulse modulation techniques and precision
  amplifiers

3000km Ericsson-Marconi system

• Amplifiers
• Typically every 80km on land, 30-40km undersea
• Festoon systems for coastal hopping systems
  offer typically 300-400 km without intermediate
  amplifiers
• Some of the technology can be common across metro
  and long haul systems - supervision,
  multiplex/demultiplex, chassis, etc

Multi-haul concept, a common platform adapted
for specific applications by changing plug-in
cards
8
Raman amplification

• Transmission fibre itself acts as a distributed
  amplifier
• Interactions1 between high energy pumped light
  and low energy signal light - span up to 400km
  if pumped from both fibre ends
• Extends the distance before an amplifier is
  needed, compared with erbium amplifiers -
  typically 80km
• Use alone or combined with conventional (erbium)
  amplifiers
• Bad news
• High pump power levels needed can burn connectors
  if not clinically clean, then needs a
  time-consuming re-termination
• Safety interlocks even more essential, for eye
  protection
• Cost - many applications dont justify the
  features
• Good news
• Ideal for island hopping or festoon systems
  around a coastline
• Create ultra-wideband optical amplifiers, at
  30-40nm per pump

9
Electronic dispersion compensation

• Fibre transport systems poor tolerance to changes
  in fibre length
• Operators rearrange fibre routes (e.g. road
  building), dont want to re-commission a working
  system
• Automatic all optical compensation is ideal but
  expensive receiver electronic compensation is
  nearly as good, and better for PMD1
• Also makes initial commissioning less critical,
  reduces cost

Why
10
Agenda

• The story so far
• How long?
• How flexible?
• Wavelength switching
• Fibre switching
• How integrated?
• How smart?
• Conclusions

11
All-optical switching

• The promise at least an order of magnitude
  smaller in size and power consumption, compared
  to electronics
• The reality complements, not displaces,
  electronics
• Simplifies provisioning and restoration of high
  capacity pipes
• Good news - excellent transparency gt almost any
  signal protocol

• Bad news - excellent transparency gt no
  visibility of data errors

• Bad news - optical technology is still mostly
  analogue, so complex networks may be harder to
  design and maintain

• Electronics development continues to compete with
  all-optical
• Electronic near-Terabit/s (640 Gbit/s) total node
  switching capacity available in compact form for
  core telecomms networks
• Comparable in total capacity with mid-sized
  all-optical switch, e.g. 64 port x 10 Gbit/s

For the highest capacity, electronic and
all-optical both have roles
12
Application areas for photonic or all-optical
switching - in the traffic path

• Generic applications
• Disaster recovery (1) and flexible signal routing
  (2), for a wide range of communication protocols
• Early emphasis was from 1998 on (2), but (1) is
  now as significant
• Types of business
• Privately owned commercial
• Switching centres for broadcast and for internet
  exchanges
• In-house data networks and production test
  facilities
• Government funded
• Defence (notably avionics), research networks
• Public networks
• Municipal optical networks - growing in US,
  Europe, Asia
• Public network operators - e.g. BT, Swisscom,
  Cable Wireless

13
All-optical switching technology options

• Many technologies demonstrated, very few
  commercialised
• Free space and planar waveguide, holograms,
  bubbles, acoustic waves, tuned lasers,
  semiconductor optical amplifiers (SOA), etc
• Two main architectures, with two main
  technologies in each
• A. ROADM (reconfigurable optical add-drop
  multiplexer)
• Wavelength blocker - usually liquid crystal
  Mostly for rings
• WSS (wavelength selective switch) - matrix of
  moving mirrors - MEMS (micro-electro mechanical
  switches), e.g. 1 x 5 each
• B. OXC (optical / photonic cross-connect) -
  switches fibres
• Matrix of moving mirror MEMs, each one gives e.g.
  32 x 32 - usually electrostatic field actuation
• Matrix of moving fibre ends, each one gives e.g.
  32 x 32 - usually piezo-electric actuation

ROADMs solution adopted by most public network
operators
14
ROADM - wavelength / lambda switching options
2
1
Split
east
west
Broadcast select or Blocker - 3 degree
Electro/optical conversion
LCD or moving mirror array
3
add drop

• One direction shown repeated for other direction
• E/O conversion or direct optical access for
  alien lambdas via policing unit

15
Trends in ROADMs

• Commercial factors
• Business case justified typically at gt 6
  wavelengths lit per fibre
• The ROADM market is the fastest growing segment
  of the resurgent optical networking market
  (Infonetics Research, Mar 06)
• 40 Gbit/s per wavelength available from some
  vendors
• ROADM pioneers from 1998 - 2002 were Marconi (now
  Ericsson) in Europe, then Fujitsu in north America

• Sequence of architectures
• 1 - thermo-electric optical switches based on
  polymer PLCs
• 2 - LCD wavelength blockers1 plus broadcast
  select for drop channels most deployed products
  use this
• 3 - wavelength selective switches (WSS) first
  products emerging
• 4 - possible future return to LCDs for
  holographic switching at high capacity

ROADMs reduce transponder capex and provisioning
opex
16
Options for all-optical fibre switching - OXC

• Photonic / optical switch / cross-connect (OXC)
  3 key suppliers
• Calient, Glimmerglass, Polatis .
• Typically 16 - 64 ports (up to 256), each with
  one wavelength
• i.e. effective capacity similar to that of a ROADM

• Disaster recovery uses the smaller versions
• Pioneers of large switches (256 ports and above)
  were Calient, Lucent, during 2000 - 2003 many
  vendors came and went
• Challenge of monitoring traffic quality remotely
  is not fully resolved
• Very few large switches deployed most business
  is for small ones

The solution adopted by most non-public-network
operators
17
Commercially successful all-optical matrix
switches move the mirror or the fibre
Mirror arrays
Fibre arrays

• MEMs mirror OXC had 38 suppliers in 2002, now
  just 21
• Typically 64-256 ports

Used in data / video switching centres, not much
in public networks
18
Agenda

• The story so far
• How long?
• How flexible?
• How integrated?
• How smart?
• Conclusions

19
Convergence trend Packet and circuit functions
in shared platforms
Services
Video
Voice
Data

• Layer 2 cards / blades also appearing in Next
  Generation SDH optical switches for the core

MSPP multi-service provisioning platform,
network or customer sites
20
Micro MSPPs spot the difference

• Micro MSPPs sit at the customer site, provide
  multiple service interfaces
• Both shown offer multiple ports of Ethernet and
  of E1
• Which is based on packets, which on TDM?
• Which looks more expensive?

• Upper is Ericsson (ex-Marconi) OMS840, based on
  Eth/GFP/TDM
• Ericsson (ex Axxessitt) AXX 9200 is another in
  this family
• Lower is from another vendor, based on packets

• Both images slightly edited to conceal identity

SDH-based Ethernet has low cost
21
Agenda

• The story so far
• How long?
• How flexible?
• How integrated?
• How smart?
• Conclusions

22
OTN - Optical Transport Network

• OTN defined by ITU-T, new hierarchical frame
  structure, etc
• Primarily for optical transport mesh in the core,
  beyond SDH
• Standardised transport and monitoring of high
  capacity payloads, to carry both packets and
  circuits
• Converges platforms and management / control
  planes, with SDH and WDM typically combined with
  SDH / WDM platform
• Misnamed as all-optical because of its relative
  transparency to payload format
• Global activity
• Support for frame structure G.709 as an interface
  is wide (routers, optical switches) support
  emerging for G.709 switching
• Interest started in Europe, growing in US
• Key roles in carriers carrier, also bypassing
  routers
• Reduces router typically high levels (60-70) of
  transit traffic - improves resilience, latency,
  etc, reduces network capex

23
Smart optical networks

• ITU-T defined G.8080 - ASON (network) and G.807 -
  ASTN (control) for automatically switched optical
  transport networks
• GMPLS emerged as preferred control technology
  with SDH/OTN
• From convergence of proprietary optical
  restoration and work on MPLS control plane
  (briefly called MP-lambda-S)
• Optical mesh reduces bandwidth for restoration by
  typically 30 compared to rings gt major capex
  savings
• ASTN control plane can be linked to data services
  control plane for faster and automated
  provisioning
• Preferred network model is hybrid of overlay
  peer, in which control planes share a limited
  view of network resources
• ASTN provides industry-standard approach to fast
  restoration
• Typical time lt 0.5 secs ok for most applications
• Simpler configuration of network resilience gt
  lower opex

GMPLS-based control reduces both capex and opex
24
Future trends

• Hardware / components
• 100 Gbit/s Ethernet - 3-5 years away
• Photonic band gap devices - smaller cheaper

• New fibres - photonic crystal fibre for extreme
  features
• Optics on silicon - mostly for embedded systems,
  rather than for extended infrastructure
• Software / systems
• Linkage between control plane for optical
  transport and control plane(s) for data switching
  gt faster provisioning
• Alternative control planes - Web 2.0 protocols /
  XML

• New systems
• Optical packet burst switching gt super-core
  networks1
• Application to GRID networking gt high capacity
  processing
• WDM PONs gt 1Gbit/s services to the home

25
Conclusions

• WDM systems scale from city to continent, while
  sharing features
• All-optical circuit switching has become
  established in many networks, notably as
  multi-wavelength optical add-drop multiplexers
• Convergence of packet and optical circuit
  technologies has created the largest optical
  market area, in multi-service provisioning
  platforms
• Intelligence in optical networks is starting to
  impact on other layers
• Reductions in opex and capex, resulting from the
  use of optical systems, have been enormous - and
  continue

26
(No Transcript)