CPqD’s SDN Activities in Optical
DWDM Terabit Networks
Juliano R. Fernandes de Oliveira, Ph.D
Optical Networking Group Leader, CPqD

2
Motivation
Triple-Play Services
Cloud Computing and Streaming Media APPs
Big Sports
Events
Pope Jonh Paul II Death (2005)
Pope Francis Presentation (2013)
http://photoblog.nbcnews.com/_news/2013/03/14/173123
16-witnessing-papal-history-changes-with-digital-age
Brazil
8.2 x
Source: Cisco 2012
1,9
EB
0,9
EB
2010 2011 2012 2013 2014 2015
Internet
Mobile
Manged IP

3
Outline (Innovation cycle at the CPqD)
5. NG-ROADM
6. SDN controller
(development)
1. CPqD  Industry
2. CPqD’s strategy (optical comm.)
3. CPqD’s optical SDN testbed
4. SDN controller (research)
1) NETCONF-modelling: YANG
2) Automatic VON instantiation
3) Control plane with PCE
4) Cognitive EDFA
5) Adaptable flexible transponder
6) Global WSS equalization
7) ADoD as NFV example

1. (CPqD  Industry) high capacity Tx/Rx innovation
112Gbps DP-QPSK
(industry)
224Gbps DP-16QAM
single carrier
(laboratorial)
448Gbps DP-16QAM
dual carrier
(laboratorial)
1Tbps
5 carriers
(laboratorial)

1. (CPqD  Industry) Network element innovation
EDFA/Raman
Amplifier
(Industry)
Optical router
(ROADM)
(Industry)
Optical
monitoring
(discrete and
integrated
photonics)

6
1. (CPqD  Industry) Control plane
• Support to ROADMS:
• Multi-degree (#2, #3, #4 ....)
• Different capabilities/architectures: C-, CD-
CDC- ...
• Multi-rate (2.5Gbs, 10Gbps, 40Gbps)
• Compliance with WSON standards
• Advanced RWA Algorithms:
• Constrained path computation
• Primary-Backup path computation
• Lightpath protection/restoration
ROADM
OADM
ROADM
ROADM
OADM
OADM
OADM
OADM
ROADM
GMPLS
GMPLS
lightpath
lightpath

7
2. CPqD’s Strategy
R&D + Proof of Concept Products
HW  FW Mec Mngt..MKT spec.
CPqD

8
Outline (Innovation cycle at the CPqD)
3. CPqD’s optical networking testbed
4. SDN controller (research)
1) NETCONF-modelling: YANG
2) Automatic VON instantiation
3) Control plane with PCE
4) Cognitive EDFA
5) Adaptable flexible transponder
6) Global WSS equalization
7) ADoD as NFV example
8) Future vision: Predictive networks
5. NG-ROADM
6. SDN controller
(development)
1. CPqD  Industry
2. CPqD’s strategy (optical comm.)
✓ ✓

9
CPqD’s research motivation (Optical networks current scenario)
Network management
(Control plane)
Network infrastructure (Data plane)
Complex Functions
(HW and SW)
Not scalable
Distributed intelligence
along HW & net. nodes
Verticalized solutions
Lack of resources share
or virtualization
Lack of optimization
(infra structure and
performance)
100Mb/
s
10Gb/s 10Gb/s 10Gb/s
100G –
100Tb/s
PROBLEMS
Proprietary

Vertically integrated
Closed and proprietary
Slow innovation
AppAppAppAppAppAppAppAppAppAppApp
Well defined horizontal segments
Open interfaces
Fast innovation
Network
OS #2
Network
OS #3
Network
OS #1 | |
Open interface
Specialized
control plane
Specialized
Hardware
Specialized
Features
Optical network
Equipments
Open interface
Amp. Transponders
Roteadores
ópticos
CPqD’s research motivation
(Optical networks current → future scenario)
SDN

11
3. CPqD’s optical networking testbed
Laboratorial Testbed
Five node flexgrid mesh network
Homemade Network
Elements
CPQD AUTONOMOUS OPTICAL NETWORK TESTBED
WSS#3
SOM
SOD
WSS#3
SOD
SOM
SOM
SOD
WSS#3
SOM
SOD
WSS#3
SOD
SOM
SOM
SOD
WSS#4
SOM
SOD
SOD SOM
OA2
OC2
IC2
OB2
IB2SOD
SOM
SOD
SOM
OCM
SOD
SOM
OCM
SOD
SOM
SOD
SOM
OCM
SOM
SOD
OCM
SOD
SOM
SOD
SOM
OCM
Intra-Node Inter-Node
Inter-Nó Intra-Nó Inter-Nó Intra-Nó
Intra-Node Inter-Node
Intra-Node Inter-Node
Switch OSC XCP
OSC Inter-Node
#1 #2 #3
OSC Intra-Node
Switch OSC XCP
OSC Inter-Node
#1 #2 #3
OSC Intra-Node
Switch OSC XCP
OSC Inter-Node
#1 #2 #3
OSC Intra-Node
Switch OSC XCP
OSC Inter-Node
#1 #2 #3
OSC Intra-Node
Switch OSC XCP
OSC Inter-Node
#1 #2 #3 #4
OSC Intra-Node
SDN Operating System
(network management)
Node Managment
Node Managment Node Managment
Node Managment
Node Managment

12
4. Optical SDN controller (research focus)
• Specialized HW
• communication network
operating system
• applications (functions
or network services)
• Communication
interfaces
• Graph network abstraction
• Legacy control plane
virtualized (GMPLS)
• infra-structure share
(spectral segmentation)
• Global network monitoring
• Adaptive, cognitive and
autonomous performance
optimization
• Transactions support
• Policies
support;
• PCE, RWA,
RSA support
• Fault
prediction
support;
Source: CPqD Globecom 2013

13
4.1) YANG (NETCONF modeling language) for Opt. Net. OS
• NETCONF-modeling language YANG
models ROADM building blocks and
its interconnections.
• The YANG model turns into a Multi-
graph Abstraction (nodes, edges)
• O-NE Concatenation through YANG
model
• Network integrated model
• Whole network analogue to a multi-
chassis NE, while an O-NE is analog
to a line-card in a chassis
Experimental
Network with 5
ROADMs
KEY:
Black nodes: Chassis system (model) ROADM Black Edges: ROADM interfaces
Red nodes: Input interfaces Red edges: Connectivity NE (ACTIVE)
Blue nodes: Output interfaces Blue edges: Connectivity NE (PASSIVE)
Orange edges: Fibers connecting OUT -> IN interfaces

14
4.2) Optical Networking (Virtualization)

15
4.2) Automatic VON instantiation
• Optical network virtualization through
spectral segmentation for each network:
• HS-VON (High Speed Virtual Optical Network):
Optical network for 100Gbps channels and
beyond;
• Aims to avoid excessive signal degradation;
• Main-VON: Legacy optical network (for signals
with rate lower than 100 Gbps);
LSP Creation
Source: CPqD Globecom 2013

16
4.3) Control plane with PCE
• Flex-Grid ROADMs:
• Lightpaths with variable spectrum widths
• GMPLS extensions (OSPF-TE, RSVPL-TE)
• PCE-based architecture:
• Centralized path computation
• Advanced RSA algorithms
• Fragmentation aware policies
• Advanced Impairments Aware algorithms
• Estimated OSNR aware paths
ROADM
ROADM
ROADM
ROADM
PCE
ROADM
GMPLS
GMPLS
100Gbps
400Gbps
Source: CPqD SBRC 2014

17
4.4) Optical Networking (Cognitive Amplifier)
• Cognitive process based on:
• Adaptative process based on optical amplifier operation points discrete performance
caracterization;
• Then a machine learning process based on neural networks is used for extrapolation of
operating region points and cognitive process feedback supported by channels bit error rate
(BER);
• Goal: channels performance maximization (lowest noise figure with good flatness);
• Result: 400% (6 dB) QoS enhancement;
Cognitive Amplifier
(Neural Networks)
Adaptive Algorithm Source: CPqD OFC/NFOEC 2013
BER
Gain

18
4.5) Adaptable Flexible Transponder
• Adaptive process:
• The QoS parameter (OSNR) is obtained through monitoring application, than in case of minimum
threshold point achieved, the transmitter modulation format is reconfigured to enhance
performance;
• Goal: Keeping Tx/Rx rate even under drastic network conditions allied to spectral efficiency
maximization;
• Result: Error free transmission for a 448Gbps signal under 22 dB OSNR degradation
• Reconfiguration: 448Gbps (2 carriers, 28Gbaud x4 symbols) 16QAM;
448Gbps (4 carriers, 28Gbaud x2 symbols) DP-QPSK;
Source: CPqD OFC/NFOEC 2014 invited & JOCN 2014

19
4.6) Global WSS Spectrum Equalization
• Global spectrum equalization of ROADMs inside an optical network:
• Na global equalization is able to optimize the whole link better than the selfish local equalization
strategy;
• Goal: Transmitted signals OSNR maximization in reconfigurable optical network link ROADMs
based;
• Result: Global equalization enhaces up to 316% (5 dB) the transmitted signals OSNR, considering
an optical DWDM link transmitting 8.96 Tbps (80x112Gbps);
80x112Gbps
(8Tbps)
Local equalization – attenuation sum
Global equalization – 1st Iteration
Global equalization – steady state
5 dB
Source: CPqD OFC/NFOEC 2014 (Accepted)

20
4.7) Add/Drop on Demand as NFV example
• Add/drop bank on demand (ADoD) consists of an optical backplane that
interconnects each degree (line interface) different modules and transponders.
Rx Rx Rx
From
degree
1
Optical Backplane
…
k×m
Rx
N 1 N
Towards
degree
Tx Tx Tx
…
k×m
Tx
…k
…ADoD modules
…k …k …k
… …
Rx Rx Rx Rx Rx
From
degree
1 3
Rx Rx Rx Rx Rx Rx
2
Rx
Backplane
cross-
connections
From
degree
TFA
Rx Rx Rx Rx Rx
1 3
Rx Rx Rx Rx Rx Rx
2
Rx Rx Rx Rx Rx Rx
1 3
Rx Rx Rx Rx Rx Rx
2
Rx
From
degree
i) ii) iii)
i) Example of a synthesized ADoD (only drop direction) with degree 3
ii) # signals from degrees 1 and 2 exceeds connectivity. Then, two modules of EDFA+splitter
and a module of tunable filter array are shared.
ii) # signals from 1 and 2 decrease (i.e. handled by direct backplane cross-connections), and
EDFA+WSS is required for incoming signals from 3.
Source: CPqD OFC/NFOEC 2014 (Accepted)

21
4.8) Future vision: Predictive Networks
• With the optical networks complexity grown, proactive fault prediction systems are the key to
enhance network availability;
• An on-line fault prediction system is based on live monitoring during algorithm execution;
• A predictive fault system uses a set of methods and models based in actual and past
monitoring data to predict faults
Online failure prediction
Symptom
monitoring
Classifiers
System
models
Time series
Analysis
Function
approximation
Graph models
Instance
models
Stochastic
Models
Machine
Learning
Bayesian
Classifiers
Fuzzy
Classifiers
Feature
Analysis
Time series
prediction
Regression
Cluster models
Failure
tracking
Co-occurence
Probability
Distribution
Estimation
Bayesian
Predictors
Non-
parametric
methods
Detect error
reporting
Pattern
recognition
Statistical
tests
Rule-based
approaches

22
Outline (Innovation cycle at the CPqD)
5. NG-ROADM
6. SDN controller
(development)
1. CPqD  Industry
2. CPqD’s strategy (optical comm.)
✓ ✓
3. CPqD’s optical networking testbed
4. SDN controller (research)
1) NETCONF-modelling: YANG
2) Automatic VON instantiation
3) Control plane with PCE
4) Cognitive EDFA
5) Adaptable flexible transponder
6) Global WSS equalization
7) ADoD as NFV example
8) Future vision: Predictive networks
✓
✓

23
5. NG-ROADM (introduction)
Coupler /
splitter
WSS
KEY:
Route and Select
Degree
Inputs
Degree
Outputs
1
N
1
N
• CPqD is currently developing a NG-ROADM
(upgradable) platform with express banks:
• Broadcast and select &Route and select
• Add/drop (with different costs per λ) banks that
support:
• Colorless: Add/drop ports are not associated to a specific
wavelength.
• Directionless: Add/drop ports are not associated to a
specific ROADM input or output port.
• Contentionless: Wavelength repetition inside the same
add/drop bank is allowed (up to ROADM degree size).
• Flexible grid: Reconfigurable spectrum slots of 12,5 GHz.
• NETCONF-modeling language YANG
is used in the development of this new platform
Towards / from
add/drop bank
Broadcast and Select
Degree
Inputs
Degree
Outputs
1
N
1
N
Towards / from
add/drop bank
Developed
Under
development

24
5. ROADM-NG (DBANK)
• Directionless Bank (DBANK) card
• WSS switching from/to express bank
• Mux/demux from/to add/drop ports
• 80 add ports, 80 drop ports
KEY:
Black nodes: Chassis system (model) NE
Black edges: NE interfaces
Red nodes: Input interfaces
Blue nodes: Output interfaces
Red edges: Connectivity NE (ACTIVE)
Coupler / splitter
WSSKEY:
Mux / demux
Demux Mux
From / towards express bank
Drops Adds
DBANK
card

25
5. ROADM-NG (CD)
• Colorless Directionless Filtered Bank (CD) card
• WSS switching from/to express bank
• WSS to drop ports, Coupler from add ports
• 40 add ports, 40 drop ports
Coupler / splitter
WSS
KEY:
From / towards express bank
Drops Adds
CDFiltered
card

26
5. ROADM-NG (CD/CDC and CR Bank)
• Colorless Directionless Contentionless Bank (CDC) card
• WSS switching from/to express bank, MCS from/to add/drop ports
• 24 add ports, 24 drop ports
• Contention Resolution Bank (CR) card
Coupler / splitter
Multicast switch
KEY:
From / towards express bank
Drops Adds
MCS MCS
MCS
WSS
CD/CDC
From / towards express bank
Splitter Splitter
CR Bank
CD/CDC
CR Bank
Contention λ’s path
Main λ’s path

27
5. Mixed add/drop banks (CDC) evolutive solution

28
5. ROADM-NG (Multilayer Control plane)
• ROADM/ODU-k switching platform:
• CPqD ROADMs
• Padtec OTN Switches
• GMPLS-based WSON solution
• PCE-based architecture
• Centralized path computation
• Constrained dynamic traffic grooming
• ODU-i ODU-j
• ODU-k  Multi-rate lightpaths
• Multi-layer path protection algorithms

29
6. Introduction to SD convergent N (SDcN)

30
6. Architecture of the solution

31
6. What are (and how to use) applications for SDcN
Applications
• Aalgorithms with a specific goal (purpose) that act on the network via instructions
allowed by the Application Server.
• Receive events from the network via callbacks.
• Act directly in virtual networks, considering the constraints defined by the network
operator.
• Work on agnostic concepts of access technologies.
• Have a complete and vast infrastructure, that contains:
• Historical measurements
• Libraries of services
• Periodic sampling of devices’ properties
• Thresholds analysis
• Topology discovery
• Devices monitoring

32
CPqD Strategy – Towards Terabit Optical Netwoks
Design and
packaging coherent
linecard critical
components
Cognitive Optical
Networks
(GMPLS/SDN,
Amps, ROADMs,
Monitoring)
Coherent
transmission
evolution towards
NxTb/s
Focus on
INDUSTRY
(Products)

33
Acknowledgements

www.cpqd.com.brThank You!