Towards Dynamic and Scalable Optical Networks

Download Report

Transcript Towards Dynamic and Scalable Optical Networks

Towards Dynamic and Scalable Optical Networks
Brian Smith
3rd May 2005
Towards Dynamic and Scalable
Optical Networks
 What is required to deliver truly dynamic optical networks ?

A Dynamic Control Plane

Technologies for Wavelength Switching
 Considerations for achieving higher data rates + capacities.

Issues with 40Gbps

Is 100Gbps achievable ?

Getting maximum spectral efficiency in a dynamic optical network.
A Dynamic Control Plane
Lightpath Control
 To deliver on-demand gigabit lightpaths, a fast and
reliable distributed control plane is required

Reliable –transport infrastructure should remain stable
during reconfigurations

Fast – otherwise its not on-demand !
 GMPLS is one control plane under development for
optical networks by the industry.

Based on standard set of IP routing and signaling protocols
 UCLP is an example of an R&E initiative (CANARIE)

TL1 interfaces controlled by a distributed service layer
based on a web browser and network model.
Evolving Towards Dynamic
Lightpath Control
 Features Required for Light-path Control

Topology discovery and link management.

Operator signaled light-paths (Network should
automatically manage its own demand).

Client on-demand light-paths (High end users can
individually control wavelengths). An important
feature for future R&E networks !

Integration with IP/MPLS control plane for
dynamic traffic engineering e.g HOPI / DRAGON.

Dynamic protection – if required
On-Demand Bandwidth Capacity
A
Server
Farm
B
Access
Network
User Controlled
Lightpath
(e.g. for
nightly data
back-up)
Wavelength
Switch
Routers
Mesh
Network
Backbone
Network
High End
Users
C
D
Available now between routers. Needs to evolve to support high
data rates on wavelengths
Example of need for on-demand
wavelengths
 4 radio telescopes in an array – 12 hour observation
 Assuming 1Gbps per telescope – 0.2 Petabits of data !
 How long would it take to back up the data to storage ?

With 100Mbps rate – ~23 days minimum assuming no packet loss.

With dedicated GigE wavelength – ~4 days.

If user can request an on demand 10GigE wavelength – ~5 hours.
High-end Research users will require high capacity on demand
services
Enabling Technologies For
Dynamic Networks
Enabling Technologies for
Dynamic Networks
 Electronic ROADM
 Optical ROADM
 Tunable lasers

Tunable 10G DWDM XFPs will be available in 2006
 Integrated optical wavelength converter / tunable laser

Demonstrated in Labs using non-linear cross-talk in Semiconductor
Optical Amplifiers – Capable of supporting up to 40Gbps
Electronic ROADM
DWDM
West
Fiber
North
Fiber
DWDM
NxN
DWDM
DWDM
Transparent
Wavelength
Switch
(electrical)
CWDM
East
Fiber
CWDM
South
Fiber
CWDM
CWDM
Trib
1310
Trib
1310
Trib
850
Trib
1550
•
Native signal transparency with layer 1 performance monitoring
•
Simple Any-to-Any Multi-Degree grid interconnection
•
Simple to Engineer.
Optical ROADM – Wave-blocker
Splitter
Wave-blocker
Drop
Filter
Coupler
Add
Filter
•
Drop and Add Filters must be tuneable for maximum flexibility.
•
Hitless filter tuning is a problem.
•
Many discrete components so expensive
•
High insertion loss – Limits DCM – Limits reach between
nodes for fully transparent networks.
Optical ROADM – Wavelength
Selective Switch (WSS)
Wavelength
Selective Switch
Drop
Channels
Coupler
Optional
Expansion
Port
Add
•
Fewer discrete optical components
•
Fully flexible colourless add/drop
•
Lower insertion loss
•
Limited number of drop ports – Use expansion port !
Comparison
- Wavelength Switching
Functionality
Electronic ROADM
Optical ROADM
Transparency
(bit rate and protocol)
Yes
- wide range of signals
Yes
Low Latency
Yes
Yes
Single wavelength granularity
(I.e. no wavelength stranding)
Yes
Yes
Mesh Support (multi-degree)
Yes
Yes
- Blocking issues
Wavelength Translation
Yes
No
Grid Conversion
(e.g. CWDM to DWDM)
Yes
No
Protocol Performance
Monitoring
Easy
Optical power only
Wavelength Protection & Hitless
Maintenance
Easy
Ring – Easy
Mesh – More difficult
Implementing ROADM Interfaces
North
Fiber
Pass-through Traffic
Optical
ROADM
I/F
DWDM
DWDM
DWDM
DWDM
DWDM
West
Fiber
South
Fiber
DWDM
NxN
DWDM
DWDM
Transparent
Wavelength
Switch
(electrical)
CWDM
CWDM
West
Optical
ROADM
I/F
CWDM
CWDM
CWDM
Trib
1310
Trib
1310
Trib
850
East
Fiber
CWDM
East
Trib
1550
Optical and Electronic ROADM complement each other.
Multi-Degree ROADM Interfaces
Optical Pass-through channels
North
Fiber
NorthW
Fiber
Optical
ROADM
I/F
Optical
ROADM
I/F
Optical
ROADM
I/F
Optical
ROADM
I/F
DWDM
DWDM
DWDM
West
Fiber
DWDM
NxN
Transparent
Wavelength
Switch
(electrical)
South
Fiber
SouthE
Fiber
DWDM
DWDM
DWDM
DWDM
East
Fiber
First step towards full NxN photonic wavelength switch.
Cost Comparison – 2.5G Traffic
COST (AU)
2.5Gbps - 2 DEGREE NODE
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
~17
1
6
11
16
21
26
# PASSTHRU WAVELENGTHS
ELECTRICAL
OPTICAL
31
36
Cost Comparison – 10G Traffic
10Gbps - 2 DEGREE NODE
2.5
COST (AU)
2
1.5
1
~6
0.5
0
1
6
11
16
21
26
# PASSTHRU WAVELENGTHS
ELECTRICAL
OPTICAL
31
36
Wavelength Switching
- Cost sweet spots
Note:
For 2-degree metro ring applications. Also
applies to 4-degree mesh architecture
Channel
Rate
10G
Electronic
ROADM
Optical
ROADM
Electronic
ROADM
2.5G
4
8
12
Optical
ROADM
16
20
24
Pass-through Channels
28
32
The Future of 40G/100G
40Gbps/100Gbps
 40Gbps

40Gbps DWDM trials and demonstrations becoming more common.

Ability to overlay on existing 2.5/10G links – a key driver !

40Gbps router interfaces have been demonstrated.

Dispersion must be controlled within ± 62 ps/nm.

PMD is an issue. Cannot exceed 2ps (outage < 3min/year)
 100Gbps

Can 100Gbps be achieved over DWDM ?

Dispersion tolerance even tighter - ± 25 ps/nm.

PMD more of an issue. Cannot exceed 1ps (outage < 3min/year)
40Gbps Dispersion Tolerance
6x80kmx26dB - 32
• 100GHz spacing SPM, XPM and FWM effects included
6
Penalty (dB)
5
4
3
2
1
0
-600
-400
-200
0
200
400
600
Net Dispersion (ps/nm)
40Gbps
10Gbps
Range of possible net
dispersion
Tunable Dispersion Compensation Required for 40Gbps.
100Gbps
 Several published examples of single wavelength 100Gbps+
transmission.
 Spectral width ~ 150 GHz for NRZ so won’t fit into a 100GHz
spaced DWDM pass-band (~85GHz) !
 Dispersion limit for NRZ is ± 25ps/nm.
 If we use non-binary coding – Spectral width reduced to 75GHz
– Just fits within 100Ghz spaced DWDM band.
 Needs tight control of laser + filter wavelengths.
 Using >1 bit per symbol coding technique such as duo-binary
or QPSK improves tolerance to dispersion and PMD.
100Gbps is achievable. Needs sophisticated coding!
Polarization Mode Dispersion
 Using 6x80kmx26dB with 6 EDFA and 6 DCM, the calculated average
DGD (assuming fiber is post 1995) = 2.5 ps
 The PMD tolerance (and expected outage) for various data rates is:
Rate
pmd tolerance
system outages/yr
2.5G
30ps
insignificant pmd outages/yr
10G
7.6ps
insignificant pmd outages/yr
40G(NRZ)
2ps
~ 3 minutes/year assuming FEC
100G(NRZ)
0.9ps
Requires PMD compensation
How Much Capacity ?
Duo-binary
100Gbps
16 symbol
levels – 4 bits
per symbol
required.
256 symbol
levels – 8 bits
per symbol
required.
NRZ/CS-RZ/
Duobinary
Wave-locker+
Wave-locker+
16 symbol
levels – 4 bits
per symbol
Wave-locker++
1b/s/Hz
40Gbps
10Gbps
10G overlay 
0.4b/s/Hz
0.8b/s/Hz
No issue
Reduced reach
Reduced reach
NRZ
Wave-locker
No ROADMs
0.1b/s/Hz
NRZ
Wave-locker+
0.2b/s/Hz
0.4b/s/Hz
50GHz
25GHz
100GHz
Summary
 On-Demand Light-path Control Enabled by:

Distributed, Intelligent Light-path Control (UCLP, GMPLS)

Electronic and Optical ROADM.

Widely Tunable Laser Sources.
 40Gbps/100Gbps

40Gbps can be deployed over existing 10G infrastructure with
appropriate dispersion control + FEC.

100Gbps will a challenge requiring sophisticated coding schemes
and components for PMD mitigation.
Thank You
Impact of Tighter Channel Spacing
 Four Wave Mixing (FWM)
FWM EFFICIENCY (dB)
0.00
-5.00
-10.00
-15.00
-20.00
-25.00
-30.00
0
20
40
60
80
100
120
WAVELENGTH SPACING (GHz)
SMF-28
LEAF
DSF
Increased FWM Impact – Reduced Reach.
Impact of Tighter Channel Spacing
XPM EFFICENCY (dB)
 Cross Phase Modulation distortion (XPM)
0
-2
-4
-6
-8
-10
-12
-14
-16
-18
-20
0
20
40
60
80
100
120
WAVELENGTH SPACING (GHz)
SMF-28
LEAF
DSF
Increased XPM Impact – Reduced Reach.