Flexible optical networking with spectral or spatial super-channels
1. 1
âNetworks and Optical Communicationsâ research group â NOC
Flexible optical networking with
spectral or spatial super-channels
Presented by: Dr. Ioannis Tomkos (itom@ait.gr)
Co-Authors: P. S. Khodashenas, J.M. Rivas-Moscoso, D. Klonidis,
D. M. Marom, G. ThouĂŠnon, A. Ellis, D. Hillerkuss, J. Zhao, D. Siracusa, F. JimĂŠnez, N. Psaila
IV International Workshop on trends in optical technologies
Campinas, Sao Paulo, Brazil â May 27th & 28th 2015
2. 2
AITâs role in the optical network evolution
Scope:
Research on architectures,
protocols, algorithms, transmission
systems and technologies for high-
speed telecommunication systems
applicable in backbone networks,
access networks and
interconnection of servers (DCNs)
and processors (HPC)
Scientific Results (2003-2015):
ďž Over 150 publications in archival
scientific journals and magazines
(including best paper awards and
highly cited papers)
ďž Over 400 publications in major
international conferences and
workshops
ďž Participated in over 25 research
projects: 5 projects within FP6 and 12
projects within FP7 and 1 H2020
ďž Led 8 EU research projects as
Technical Manager of the entire
consortium: 2 FP6 and 6 FP7
4. 4
Presentation overview
ď§ Evolution of Optical Communication Systems & Networks
ď§ Spectrally flexible optical networking
⢠The activities of EU project FOX-C
ďŽSpectrally flexible super-channel transceivers
ďŽNodes for all-optical add/drop of sub-channels
ďŽNetworking studies to demonstrate the benefits of FOX-C solution
ď§ Spatially flexible optical networking
⢠The activities of EU project INSPACE
ďŽNodes for independent or joint switching of SDM super-channels
ďŽNetworking studies to demonstrate the benefits of INSPACE solution
ďŽDevelopment of an SDN-based control plane for SDM-based networks
ď§ Summary & Conclusions
5. 5
Historical evolution of optical communications system
capacity and bit-rate ď´ distance product
0.00001
0.0001
0.001
0.01
0.1
1
10
100
1000
0
1
10
100
1,000
10,000
100,000
1,000,000
10,000,000
1983 1987 1991 1995 1999 2003 2007 2011 2015
TotalFibreCapacity(Tbit/s)
BitRateDistanceProduct
(Gbit/s.Mm)
Year Published
WDM
TDM
OFDM/CoWDM
Coherent Detection
Spatial Multiplexing
Total capacity
⢠Traffic increases at a rate of 20-40%
per year, while capacity of deployed
SMF-based networks approaches
fundamental limitsâŚ
⢠New traffic characteristics lead to
new network requirements:
⢠Rapidly changing traffic patterns
⢠High peak-to-average traffic ratio
⢠Ultra-large data-chunks transfers
⢠Asymmetric traffic between nodes
⢠Increasing high-QoS traffic
⢠Fiber bandwidth was consider for many years as an abundant resource, but we have
almost utilized to the maximum extend the EDFA amplifiers bandwidth (i.e. while
approaching the fundamental SE limits)
⢠A short-term solution is to utilize the available fiber spectrum more efficiently/wisely as is
the case in wireless networks where bandwidth was always a limited/scarce resource -
(Spectrally flexible systems/networks)
⢠A forward-looking option is to deploy new fibers (or use strands of available SMF fibers)
that can support multi-cores or/and multi-modes per core (SDM/Spatially-flexible systems/
networks)
Data from Prof. Andrew Ellis
7. 7
Main building blocks to enable spectrally
flexible optical networking
Flexible Optical Networking
Flexible
transponders
Strong research
field over last 5
years
Network planning
and control plane
issues
Networking
studies have proved
the benefits of flexible
networking
Flexible switching
nodes
Limited number of
demos showing
mostly âdropâ
function
Ioannis Tomkos et. al., âA Tutorial on the Flexible Optical Networking Paradigm: State-of-the-Art, Trends, and
Research Challengesâ, Invited paper at the âProceedings of the IEEEâ (Impact Factor: 6.91) 05/2014
9. 9
How to add/drop sub-channels out of super-channels
in a three-level spectrally flexible optical node?
ď§ Level 3 ď Express through or add/drop of Tb super-channels (via conventional WSSs)
ď§ Level 2 ď Processing (add/drop/erase) of super-channel contents
⢠Offers grooming capabilities in the optical domain! â How to do it?
ď§ Level 1 ď Generation/Detection and regeneration of sub-channel contents
⢠Electronic processing
10. 10
The FOX-C project consortium
ď§ Optronics Technologies S.A
⢠Mr. George Papastergiou (Coordinator)
⢠Dr. Marianna Angelou
⢠Dr. Thanasis Theocharidis
ď§ Finisar Israel LTD ⢠Dr. Shalva Ben-Ezra
ď§ W-Onesys S.L. ⢠Dr. Jordi FerrĂŠ Ferran (WP6 Leader)
ď§ Orange Labs â FT
⢠Dr. Erwan Pincemin (WP5 Leader)
⢠Dr. Christophe Betoule
⢠Dr. Gilles Thouenon
ď§ Athens Information Technology
ď§ The Hebrew University of
Jerusalem
ď§ EidgenĂśssische Technische
Hochschule ZĂźrich
ď§ University College Cork
ď§ Aston University
⢠Dr. Ioannis Tomkos (Technical Mngr)
⢠Dr. Dimitrios Klonidis (WP2 Leader)
⢠Dr. Pouria S. Khodashenas
⢠Dr. JosÊ M. Rivas-Moscoso
⢠Prof. Dan Marom (WP4 Leader)
⢠Prof. Juerg Leuthold
⢠Dr. David Hillerkuss (WP3 Leader)
⢠Mr. Benedikt Bäuerle
⢠Dr. Jian Zhao
⢠Prof. Andrew Ellis
⢠Dr. Stylianos Sygletos
⢠Dr. Simon Fabbri
⢠Dr. Andreas Perentos
14. 14
The FOX-C node architecture for N-WDM super-channels
Enables all-optical add/drop of sub-channels out of
non-spectrally-overlapping (q)N-WDM super-channels
15. 15
FOX-Câs novel ultra-fine spectral resolution filters
ď§ Based on a state-of-the-art phased array
implemented with a high resolution AWG
ď§ Achieved record resolution and addressability
values
⢠Record resolution: <1GHz
⢠Record addressability (spectral granularity): 200MHz
* Roy Rudnick, et. al., âOne GHz Resolution Arrayed Waveguide Grating Filter with LCoS Phase Compensationâ,
in proc. OFC/NFOEC 2014, paper Th3F.7
16. 16
The EU project FOX-C node architecture
for AO-OFDM super-channels
Enables all-optical add/drop of sub-channels out of
spectrally-overlapping OFDM super-channels
17. 17
Novel all optical ROADM for OFDM signals
* S. Sygletos et al., âA Novel Architecture for All-Optical Add-Drop Multiplexing of
OFDM Signalsâ, in proc. ECOC 2014, Sept. 2014.
18. 18
ď§ Results demonstrate that, in the case of qN-WDM, there is negligible reach
penalty when the FOX-C nodes are considered along the signal path
Transmission studies with cascaded FOADMs
N-WDM qN-WDM
ďf = 25 GHz ďf = 25.9 GHz ďf = 26.8 GHz
19. 19
Nyquist WDM super-channel composed of Nyquist-
shaped sub-channels: Do we need âGridlessâ?
12.5GHz
187.5 GHz
200 GHz
134 GHz
150 GHz
(*) P. Sayyad Khodashenas et al., âEvaluating the Optimum Filter Resolution and Sub-Channel Spectrum Granularity
for Flexible Super-Channelsâ, OFC 2015, paper W1I.5.
ď§ Sub-band allocation options according to frequency slot width
⢠ITU-T 12.5 GHz grid
⢠Gridless
The super-channel bandwidth depends on the chosen sub-channel granularity.
Two examples are shown above:
ďź Super-channels allocated on ITU-T 12.5 GHz grid (including GB = 12.5GHz).
ďź âGridlessâ operation â Does it offers significant advantages?
* P. S. Khodashenas, J. M. Rivas-Moscoso, D. Klonidis, D. M. Marom and I. Tomkos, âEvaluating the Optimum Filter
Resolution and Sub-Channel Spectrum Granularity for Flexible Super-Channelsâ, in proc. OFC/NFOEC 2015, paper W1I.5.
20. 20
Networking studies to derive specifications
ď§ Optimized sub-channel slot size from a network-level perspective:
⢠Flex-grid qN-WDM systems with frequency-slot size of 12.5 GHz and coarse
switching in GĂANT2 pan-EU network topology.
⢠Optimum sub-channel grid was investigated:
Best
compromise
* P. S. Khodashenas, J. M. Rivas-Moscoso, D. Klonidis, D. M. Marom and I. Tomkos, âEvaluating the Optimum Filter
Resolution and Sub-Channel Spectrum Granularity for Flexible Super-Channelsâ, in proc. OFC/NFOEC 2015, paper W1I.5.
21. 21
Techno-economic studies â FOX-C vs legacy solutions
ď§ Is the FOX-C solution worth considering for real deployments?
⢠Are the resulting network-wide capital expenditure savings significant
enough to justify a FOX-C-like solution?
⢠Inputs for the analysis:
ďŽNetwork topology, traffic matrix (FT/Orange national network)
ďŽCost model *
ďŽIt requires also a novel routing, modulation level and spectrum allocation algorithm that
matches the FOX-C networks solution characteristics (AOTG-RMLSA) **
⢠Outputs from the analysis:
ďŽUtilized resources (such as transceivers, nodes and spectrum) to guarantee blocking-free
connection establishment, while minimizing the spectral occupancy.
â Benchmarks:
Âť SLR over fixed-grid (widely deployed network solution)
Âť MLR over flexi-grid (common understanding of flexible optical networks)
* Ref: J. Rivas-Moscoso, S. Ben-Ezra, P. Khodashenas, D. Marom, D. Klonidis, P. Zakynthinos & I. Tomkos âCost & Power
Consumption Model for Super-channel Transmission with All-Optical Sub-channel Add/Drop Capabilityâ ,(Invited) ICTON 2015.
** Ref: P. S. Khodashenas, J. M. Rivas-Moscoso, D. Klonidis, G. ThouĂŠnon, C. Betoule and I. Tomkos, âImpairment-aware
Resource Allocation over Flexi-grid Network with All-Optical Add/Drop Capabilityâ, submitted to ECOC 2015.
22. 22
FOX-C cost & power consumption model
ď§ Cost and power consumption model:
⢠Benchmark: Single carrier 100G transceiver
⢠Tb/s super-channel transceiver based on:
ďŽ Electrical multiplexing schemes:
â NFDM, NWDM with electrical filtering, MB-e(f)OFDM
ďŽ Optical multiplexing schemes:
â (q)NWDM with optical filtering
â Conventional AO-OFDM with DSP
⢠ROADM implementations:
ďŽSupporting non-overlapping sub-channel A/D
ďŽSupporting overlapping and non-overlapping sub-channel A/D
⢠Sensitivity analysis
J. Rivas-Moscoso, S. Ben-Ezra, P. Khodashenas, D. Marom, D. Klonidis, P. Zakynthinos & I. Tomkos âCost & Power
Consumption Model for Super-channel Transmission with All-Optical Sub-channel Add/Drop Capabilityâ ,(Invited) ICTON 2015.
23. 23
Benchmark: 100G transceiver
ď§ SLR/MLR 100G transceiver:
DSPchip
Datain/out
ď°/2
Q I
ď°/2
DP IQ Mod
Q I
DAC
DAC
DAC
DAC
ADC
ADC
ADC
ADC
ECL
LO
Drivers
RF LP filters
RF LP filters (optional)
PBS
PBS
Sx
Sy
LOy
LOx
Ix
Qx
Iy
Qy
Balanced
photodetector
DFB
Intensity
modulatorsplitter
50/50
211
-1PRBS
(3.5or7)
Gb/s
BPG
4:1
MUX
QPSKâ16QAM
to
Balanced
photodetector
DFB
Intensity
modulatorsplitter
50/50
211
-1PRBS
(3.5or7)
Gb/s
BPG
4:1
MUX
QPSKâ16QAM
to
Balanced
photodetector
DFB
Intensity
modulatorsplitter
50/50
211
-1PRBS
(3.5or7)
Gb/s
BPG
4:1
MUX
QPSKâ16QAM
to
Balanced
photodetector
DFB
Intensity
modulatorsplitter
50/50
211
-1PRBS
(3.5or7)
Gb/s
BPG
4:1
MUX
QPSKâ16QAM
to
90Âş Hybrid
TIA
DP coherent receiver
24. 24
Benchmark: 100G transceiver
ď§ SLR/MLR transceiver:
(*) Relative to cost of 100G transceiver.
(**) Relative to cost of 10G transceiver
SLR/MLR TRx
Component
Relative Unit
cost (*)
Power (W)
[max]
#
Relative
cost (*)
Relative
cost (**)
Total
power (W)
DSP Chip 0.36 38.5 1 0.36 1.9 38.5
PM IQ Mod 0.22 0.0 1 0.22 1.1 0.0
Laser (Tx & Rx LO) 0.05 1.5 2 0.11 0.3 3.0
4-Port Modulator Driver 0.07 6.0 1 0.07 0.4 6.0
RF LP filter 0.004 0.0 8 0.03 0.0 0.0
DP Coherent Receiver 0.22 1.5 1 0.22 1.1 1.5
1.00 5.2 49.0
Actual cost (not
price!) in the range
of 25-30K!!!
25. 25
Cost/power comparison of spectrally
flexible super-channel transceivers
ď§ Cost/power per sub-channel for super-channel transceivers
capable of generating different numbers of sub-channels:
(*) Relative to cost of 100G transceiver.
Electrical
multiplexing
schemes
NWDM with
optical filters
AO-OFDM
Number of
sub-
channels
Cost (*) P (W) Cost (*) P (W) Cost (*) P (W)
4 0.80 50.8 0.98 52.8 0.79 50.8
6 0.76 49.2 0.84 50.5 0.75 49.2
8 0.74 48.4 0.77 49.4 0.73 48.4
10 0.73 47.9 0.73 48.7 0.72 47.9
12 0.72 47.6 0.71 48.3 0.71 47.6
26. 26
Non-overlapping sub-channel A/D capable OXC node
ď§ Node with D=3 degrees:
W Sch Tx
NRxSRx
N Sch Tx
S Sch TxW Rx
Switches with
D-1 switching
states
Sbch Tx
Sbch Tx
Drop Add
Sbch
Tx
W
S
N
A/D
+22
+10
+22+10
+22+10
-5
-5
-5 -5
-5 -5
+8 +14 -0.5 -0.5-15
-22 dBm/50 GHz
-2 dBm/C-band
0
20
-5 / 15
-5
-10 / 10
0 / 20
0
-15
-10
0
-5
Numbers in
black: dB
Numbers in
green: dBm
A/D
A/D
Number of sub-channel
add/drop cards M = 3
A/D card
based on
HSR filter in
R. Rudnick,
ECOC 2014,
PD.4.1
27. 27
Non-overlapping sub-channel A/D capable OXC node
ď§ Node with D=3 degrees and M = 3 A/D cards:
(*) Cost relative to 100G transceiver cost
(**) Cost relative to 10G transceiver cost
(***) Cost relative to cost of an OXC node of degree D without sub-channel A/D capability
(^) Due to sharing of management between amplification modules. Factor is applied to number of amplifiers minus 1
F-OXC with degree D and M HSR filters
Component
Relative
unit cost (*)
Power
(W)
#
Relative
cost (*)
Relative
cost (**)
Relative
cost (***)
Power
reduction
factor (^)
Total Power
(W)
1Ă20 WSS 0.54 4.00 6 3.24 16.92 0.86 24.0
1x1 HSR filter 0.54 4.00 3 1.62 8.46 0.43 12.0
Variable gain dual-stage
amplifier
0.18 12.00 6 1.08 5.64 0.29 0.10 66.0
1x(D-1) switch 0.01 0.00 6 0.09 0.45 0.02 0.0
6.02 31.47 1.59 102.0
Note: A similar investigation was performed for nodes suitable for overlapping sub-channels
28. 28
Overlapping sub-channel A/D capable OXC node
ď§ Node with D=3 and M=3 for an A/D card implementation based on N gates
F-OXC with degree D and M TIDE filters
TIDE with N gates
Component
Unit
cost (*)
Power
(W)
#
Cost
(*)
Cost
(**)
Cost
(***)
Power
reduct.
factor
(^)
Total
Power
(W)
1Ă20 WSS 0.54 4.00 6 3.24 16.9 0.9 24.0
1x1 HSR filter 0.54 4.00 0 0.00 0.0 0.0 0.0
1xN HSR filter 0.72 4.00 4 2.88 15.0 0.8 16.0
Integrated TIDE 0.11 4.00 2 0.22 1.1 0.1 8.0
Variable gain
amplifier
0.11 9.00 2 0.22 1.1 0.1 0.10 16.2
Variable gain dual-
stage amplifier
0.18 12.00 5 0.90 4.7 0.2 0.10 55.2
1x(D-1) switch 0.01 0.00 4 0.06 0.3 0.0 0.0
7.50 39.2 2.0 119.4
(*) Cost relative to 100G transceiver cost
(**) Cost relative to 10G transceiver cost
(***) Cost relative to cost of an OXC node of degree D without sub-channel A/D capability
(^) Due to sharing of management between amplification modules.
29. 29
Results of techno-economic studies comparing
FOX-C vs SLR/MLR legacy solutions
* P. S. Khodashenas, J. M. Rivas-Moscoso, D. Klonidis, G. ThouĂŠnon, C. Betoule, E. Pincemin and I. Tomkos, âTechno-
Economic Analysis of Flexi-Grid Networks with All-Optical Add/Drop Capabilityâ, submitted to PS2015.
~15%
~30%
~30%
FOX-C based solutions can
offer up to 30% cost savings
compared to non-grooming
capable end-to-end solutions
using either SLR or MLR
30. 30
Techno-economic studies â FOX-C (all-optical grooming)
vs. OTN (electronic grooming ) based solutions - I
* G. ThouĂŠnon, C. Betoule, E. Pincemin, P.S. Khodashenas, J.M. Rivas-Moscoso, I. Tomkos, submitted to ECOC 2015.
S0: SLR over fixed-grid
S1: Nyquist WDM
S2: MB-OFDM
⌠but can FOX-C based solutions offer significant cost savings
compared to conventional OTN based grooming-capable solutions?
(study performed in collaboration with France Telecom/Orange)
31. 31
Techno-economic studies â FOX-C (all-optical grooming)
vs. OTN (electronic grooming ) based solutions - II
+37% +36%
+60%
-29%
(a) For Traffic Volume V1
+23%
+10%
-21%-18%
(b) For Traffic Volume V2
V1: 7 Tbps of ingress traffic
V2: traffic increase projection spanning roughly eight years with a constant per-year traffic
growth of 35%
Global multi-layer transport network cost comparison
33. 33
Whatâs next in capacity expansion⌠In Space
ď§ Space is the obvious yet unexplored (until 2009) dimension
⢠âŚBUT by simply increasing the number of systems, the cost and power consumption also
increase linearly!
ď§ Efficient use of the space-domain requires âspatial integration of elementsâ*
⢠Significant efforts in the development of FMF and MCF (fibre integration)
⢠Multi-link amplification systems have also be proposed and developed
⢠Tx/Rx integration is a hot and very active topic
⢠Optical switches are largely unexplored so far (INSPACE focus!)
MC/FM EDFA/
EDFA array
MCF/FMF/
Bundle of SMF
Tx PIC Rx PIC
* Peter J. Winzer, âSpatial Multiplexing: The next frontier in network capacity scalingâ, Tutorial paper at ECOC 2013
34. 34
Degrees of freedom in SDM transmission/switching are defined by
the type of transmission medium and how crosstalk is handled
ď§ Core count
ď§ Mode count
ď§ Cladding diameter
ď§ Core layout
⢠Geometry
⢠Homo-/heterogeneous core
structure
ď§ Refractive-index profile
⢠Graded-index
⢠Step-index
⢠Trench-assisted
Becarefulwith:
dBUFFER=250ďm
dCLADDING=125ďm
dCORE=8ďm
SMF
ď§ Inter-core crosstalk
ď§ Inter-mode crosstalk
ď§ Differential mode group
delay (DMGD)
ď§ Bend loss
ď§ Nonlinearity
ď§ Process variability
35. 35
Degrees of freedom in SDM transmission
ď§ Core count
ď§ Mode count
ď§ Cladding diameter
ď§ Core layout
⢠Geometry
⢠Homo-/heterogeneous core
structure
ď§ Refractive-index profile
⢠Graded-index
⢠Step-index
⢠Trench-assisted
ď§ Inter-core crosstalk
ď§ Inter-mode crosstalk
ď§ Differential mode group
delay (DMGD)
ď§ Bend loss
ď§ Nonlinearity
ď§ Process variability
Becarefulwith:
dBUFFER=250ďm
dCLADDING=125ďm
dCORE=8ďm
SMF
36. 36
Degrees of freedom in SDM transmission
ď§ Core count
ď§ Mode count
ď§ Cladding diameter
ď§ Core layout
⢠Geometry
⢠Homo-/heterogeneous core
structure
ď§ Refractive-index profile
⢠Graded-index
⢠Step-index
⢠Trench-assisted
ď§ Inter-core crosstalk
ď§ Inter-mode crosstalk
ď§ Differential mode group
delay (DMGD)
ď§ Bend loss
ď§ Nonlinearity
ď§ Process variability
Becarefulwith:
Bundle of SMF
(A) Uncoupled spatial modes
37. 37
Degrees of freedom in SDM transmission
ď§ Core count
ď§ Mode count
ď§ Cladding diameter
ď§ Core layout
⢠Geometry
⢠Homo-/heterogeneous core
structure
ď§ Refractive-index profile
⢠Graded-index
⢠Step-index
⢠Trench-assisted
ď§ Inter-core crosstalk
ď§ Inter-mode crosstalk
ď§ Differential mode group
delay (DMGD)
ď§ Bend loss
ď§ Nonlinearity
ď§ Process variability
Becarefulwith:
Bundle of SMFMCF
(A) Uncoupled spatial modes
41. 41
Hero transmission experiments based on SDM
⌠BUT all these are very good for the spatial capacity increase in Point-to-Point
systemsâŚ
âŚWHAT ABOUT using the spatial dimension for optical networking
42. 42
Evolution from spectrum flexible to spatially (& spectrum)
flexible optical networking
Spectrum based
BW allocation
Spatial & Spectrum based
BW allocation
Spectrum Flexible Optical Networking
- Combined selection of channel
bandwidth (format/ data rate) and
spectral allocation according to:
demand, distance and required
performance
- Îť + format/rate tunable TxRx
- Flexible switching of variable
spectral slots at different wavelengths
- Optimized spectral usage
Spatially and Spectrally Flexible Optical
Networking
- Extend flexibility to the space switching
domain
- Multi-dimensional switching granularity
- Channel allocation over
a. multiple Modes/Cores/fibres
b. multiple spectral slots
- Optimized system bandwidth usage
- Combined spectral â spatial optimization.
- Multi-dimensional flexible switching
43. 43
The INSPACE project consortium
ď§ Optronics Technologies S.A
⢠Mr. George Papastergiou (Coordinator)
⢠Dr. Nina Christodoulia
⢠Dr. Thanasis Theocharidis
ď§ TelefĂłnica InvestigaciĂłn y
Desarrollo SA
⢠Mr. Felipe JimÊnez-Arribas (WP2 Leader)
⢠Dr. VĂctor LĂłpez
⢠Dr. Ăscar GonzĂĄlez de Dios
ď§ The Hebrew University of Jerusalem ⢠Prof. Dan Marom (WP4 Leader)
⢠Dr. Miri Blau
ď§ Athens Information Technology
⢠Dr. Ioannis Tomkos (Technical Mngr)
⢠Dr. Dimitrios Klonidis (WP6 Leader)
⢠Dr. Pouria S. Khodashenas
⢠Dr. JosÊ M. Rivas-Moscoso
ď§ Optoscribe Ltd.
ď§ CREATE-NET (Center for Research and
Telecommunication Experimentation for
Networked Communities)
ď§ Aston University
ď§ Finisar Israel Ltd.
ď§ W-ONE SYS SL
⢠Dr. Nicholas Psaila
⢠Dr. John MacDonald
⢠Dr. Paul Mitchell
⢠Dr. Domenico Siracusa (WP5 Leader)
⢠Dr. Federico Pederzolli
⢠Dr. Elio Salvadori
⢠Prof. Andrew Ellis (WP3 Leader)
⢠Dr. Stylianos Sygletos
⢠Dr. Naoise Mac Suibhne
⢠Dr. Filipe Ferreira
⢠Dr. Christian Sånchez-Costa
⢠Dr. Shalva Ben-Ezra
⢠Dr. Jordi FerrÊ Ferran (WP7 Leader)
⢠Dr. Jaume MarinÊ
44. 44
INSPACE project channel allocation concept
Modes/Cores
Wavelengths
Data rate
(Modulation level)
Degrees of Flexibility
Modes
or
Cores
f
f
f
f
f
⢠Channels with flexible capacity can be allocated over:
â one or few modes/multi cores
â occupying a single or multiple spectral slots
: end-to-end allocated channel
âSpatial expansion of
the spectrum over
multiple modes/cores
and therefore definition
of a superchannel over
two dimensions
(instead of the
spectrum only
dimension)â
SMF-
Bundle
or
FMF
or
MCF
Frequency Frequency
ConventionalopticalOFDM OpticalfastOFDM
(N-1)/T
(N-1)/2T
Nisthechannelnumber(=7inthisexample)
(a) (b)
Frequency Frequency
ConventionalopticalOFDM OpticalfastOFDM
(N-1)/T
(N-1)/2T
Nisthechannelnumber(=7inthisexample)
(a) (b)
N-WDM
or
OFDM
or
SC-M-QAM
Fibre,
Mode
,Core
45. 45
SDM switching classification
Independent spatial/spectral channel switching Spectral channel switching
Spatial channel switching Spectral channel switching of spatial subgroups
* D. M. Marom et al.,''Switching Solutions for WDM-SDM Optical Networks'', IEEE Comm. Mag. 53, 60-68 (2015)
46. 46
SDM switching classification
Independent spatial/spectral channel switching Spectral channel switching
Spatial channel switching Spectral channel switching of spatial subgroups
(A)
R&S node
design for
independent
spatial/spectral
channel
switching
* D. M. Marom et al.,''Switching Solutions for WDM-SDM Optical Networks'', IEEE Comm. Mag. 53, 60-68 (2015)
47. 47
SDM switching classification
Independent spatial/spectral channel switching Spectral channel switching
Spatial channel switching Spectral channel switching of spatial subgroups
R&S node design for spectral channel switching across all spatial modes
(A)
R&S node
design for
independent
spatial/spectral
channel
switching
* D. M. Marom et al.,''Switching Solutions for WDM-SDM Optical Networks'', IEEE Comm. Mag. 53, 60-68 (2015)
48. 48
SDM switching classification
Independent spatial/spectral channel switching Spectral channel switching
Spatial channel switching Spectral channel switching of spatial subgroups
R&S node design for spectral channel switching across all spatial modes
(C)
OXC design for spatial channel switching across all spectral channels
* D. M. Marom et al.,''Switching Solutions for WDM-SDM Optical Networks'', IEEE Comm. Mag. 53, 60-68 (2015)
49. 49
* D. M. Marom et al.,''Switching Solutions for WDM-SDM Optical Networks'', IEEE Comm. Mag. 53, 60-68 (2015)
SDM switching classification
Independent spatial/spectral channel switching Spectral channel switching
Spatial channel switching Spectral channel switching of spatial subgroups
(C)
OXC design for spatial channel switching across all spectral channels
(D)
R&S node design for hybrid fractional space-full spectrum switching granularity
50. 50
Comparison of SDM switching options
Space-wavelength granularity Space granularity Wavelength granularity
Fractional space-full
wavelength granularity
Minimum
switching
granularity
Bandwidth of a single WDM channel
present at a single spatial mode.
Bandwidth of entire optical
communication band carried on a
single spatial mode.
Bandwidth of a single WDM
channel spanning over all spatial
modes.
Bandwidth of a single
WDM channel over a
subset of spatial modes.
Realization
With OXC: High-port count OXC and at
least 2M conventional WSS per I/O fiber
link.
Without OXC: 2M conventional WSS per
I/O fiber link. 4M if WSS placed on
add/drop.
Moderate port count OXC, and 2 WSS
per mode selected for WDM channel
add/drop.
4 joint switching WSS per I/O
fiber link in route-and-select
topology applied to all spatial
modes in parallel.
4ĂM/P joint switching
WSS modules per I/O fiber
link.
Flexibility
With OXC: Each mode/WDM channel
independent provisioned and routed.
Supports SDM lane change. Single point of
failure.
Without OXC: Each mode/WDM channel
independently provisioned and routed.
Spatial mode maintained. Prone to
wavelength contention.
The complete optical communication
band is routed across network. Coarse
granularity. If WDM channels need to
be extracted from many modes then
WSS count quickly escalates.
Each spatial super-channel
provisioned across all modes.
Susceptible to wavelength
contention. Add/drop bound to
direction.
Compromise solution using
small SDM groups. More
efficient when provisioning
low capacity demands.
Scaling
With OXC: Can quickly escalate to very
large port counts.
Switching node cost linearly scales with
capacity, no price benefit to SDM.
Conventional OXC can support
foreseeable mode and fiber counts.
OXC is single point of failure. Pricing
favorable but with greater add/drop
require more WSS modules.
Cost roughly independent of SDM
count. Inefficient for low capacity
connections due to minimum BW
provisioned across SDM. Large
SDM Rx/Tx are integration and
DSP challenge.
Cost scales as group count.
Groups can be turned on as
capacity grows, offering
pay-as-you-go alternative.
Maintaining small group
sizes facilitates MIMO
processing at Rx.
Estimated
loss
With OXC: 13 dB per I/O fiber link.
Without OXC: 10 dB per I/O fiber link
For MCF or FMF transmission fiber, extra 4
dB loss is induced by the spatial
MUX/DEMUX
3 dB per I/O fiber link being switched.
If add/drop from SDM fiber is
extracted, 10 dB excess loss for
through.
For MCF or FMF transmission fiber,
extra 4 dB loss is induced by the
spatial MUX/DEMUX
10 dB per I/O fiber link.
For MCF or FMF transmission
fiber, extra 4 dB loss is induced by
the spatial MUX/DEMUX
10 dB per I/O fiber link.
For MCF or FMF
transmission fiber, extra 4
dB loss is induced by the
spatial MUX/DEMUX
51. 51
SDM technology elements
ď§ INSPACE SDM Wavelength Selective Switch
⢠High port count WSS for joint switching of spatial modes
A conventional 1ď´20 WSS can turn into a 7-modeď´(1ď´2) spatial-spectral WSS.
First demonstration in OFC 2012
New port definition: Sď´(Mď´N)
S = nÂş of spatial modes
In1
Out1
Out2 M = nÂş of input fibre subgroups
N = nÂş of output fibre subgroups
52. 52
SDM technology elements
A conventional 1ď´20 WSS turns into a 7-modeď´(1ď´2) spatial-spectral WSS. First
demonstration in OFC 2012
New port definition: Sď´(Mď´N)
ď§ INSPACE SDM Wavelength Selective Switch
⢠High port count WSS for joint switching of spatial modes
ďŽBy adding a 2-D SMF array, a higher port count can be achieved
ďŽWith a fibre array of 3ď´16 (functional) fibres, a 3-modeď´(1ď´15)
spatial spectral high port count WSS has been designed/fabricated
S modes per input/output
M = 1 input
N outputs
2-D Fibre array
53. 53
SDM technology elements
ď§ INSPACE Mode MUX/DEMUX
⢠MCF breakout designed and fabricated for MCF
⢠FMF photonic lantern designed and fabricated
ďŽFabrication optimisation yielded
low IL (2 dB) with a loss uniformity
of 0.8 dB
54. 54
SDM technology elements
ď§ INSPACE Mode MUX/DEMUX
⢠MCF breakout designed and fabricated for MCF
⢠FMF photonic lantern designed and fabricated
ďŽFabrication optimisation yielded
low IL (2 dB) with a loss uniformity
of 0.8 dB
The performance of the photonic lantern is
better than competing commercial devices
and fully packaged devices are ready to be
deployed. These will be launched as an
improved product at ECOC in Sept. 2015.
55. 55
ď§ Comparison of spectral and spatial super-channel allocation
policies for SDM network operation, taking into account the
spectral efficiency/reach trade-off
ď§ First study carried out for SDM networks based on SMF
bundles
ď§ For such an SDM system, the focus is
on the comparison between two
extreme allocation strategies:
⢠Parallel systems with spectral super-channels (SpeF)
⢠Parallel systems with spatial super-channels (SpaF)
SDM resource allocation issues
⢠MCFs and FMFs with coupled transmission
cores/modes present special challenges in
terms of their physical layer performance and
implementation complexity
56. 56
ď§ SDM allocation options:
⢠A: SpeF â Spectral super-channels with flexi-grid
⢠B: SpaF â Spatial super-channels with fixed spectral width
SDM allocation options considered
57. 57
ď§ Super-channel allocation options:
⢠A: Over spectrum (SpeF)
⢠B: Over space (SpaF)
Resource allocation options and trade-offs
Big enough spacing to neglect
the crosstalk between adjacent
super-channels
* The GB size is the same for both cases
No
crosstalk
among
cores
Crosstalk between adjacent
sub-channels leads to optical
reach reduction
58. 58
Blocking results (under independent switching)
1.E-4
1.E-3
1.E-2
1.E-1
300 600 900 1200 1500
BP
Input Load [Erlang]
SpeF-Var
SpeF-34.375
SpeF-37.5
SpeF-WDM
SpaF-WDM
ď§ BP vs input load to the network for several SpeF and SpaF allocation options
(simulations performed for TelefĂłnicaâs Spain national network):
(a) SpeF-Var: SpeF using
variable spacing adapted to
the path length
(b) SpeF-34.375, SpeF-37.5,
SpeF-50: SpeF using fixed
spacing (34.375, 37.5 GHz
and 50 GHz) with 12.5-GHz
GB on both sides of each Sp-
Ch
(c) SpeF-WDM and SpaF-
WDM: SpeF and SpaF on
fixed-grid WDM conditions
with 50-GHz channel
spacing including GB.
* D. Siracusa, F. Pederzolli, P. S. Khodashenas, J. M. Rivas-Moscoso, D. Klonidis, E. Salvadori, I. Tomkos, âSpectral vs.
Spatial Super-Channel Allocation in SDM Networks under Independent and Joint Switching Paradigmsâ, ECOC 2015.
59. 59
Blocking results (under independent, joint and
fractional joint switching)
ď§ BP vs. input load to the network for several SpaF allocation
options and switching paradigms:
⢠Joint switching imposes a BP penalty compared to independent
switching, which can be minimised through proper traffic engineering
(better match between traffic profile and Tx maximum capacity)
1.E-4
1.E-3
1.E-2
1.E-1
200 400 600 800 1000 1200 1400
BP
Input Load [Erlang]
SpaF-InS
SpaF-FJoS
SpaF-JoS
60. 60
Number of WSSs required (under joint and fractional
joint switching)
ď§ Joint switching can alleviate the cost problem associated with independent
switching (resulting from the requirement of one WSS per fiber and degree) by
allowing WSS-sharing between fibers.
ď§ Total number of WSSs required, under different switching paradigms, for a
colorless, directionless R&S ROADM architecture in the Telefonica Spain
national network:
Switching Number of WSS (general)
Number of WSS
(TelefĂłnica topology)
InS 2¡Nd¡D¡S+4¡NdA/D¡S 2502
JoS 2¡Nd¡D+4¡NdA/D 278
FJoS 2¡Nd¡D¡ďŠS/Gďš+4¡NdA/D¡ďŠS/Gďš 834
ďŠÂˇďš: ceiling
S: number of fibers
Nd: total number of nodes
NdA/D: number of nodes with
A/D
D: avg. nodal degree
G: number of groups of
spatial modes (G = 3)
(For Nd = 30, NdA/D = 14, D = 3.7, S = 9, G = 3)
61. 61
Characteristic Distributed GMPLS Hybrid PCE/GMPLS Centralized SDN
Implementation
complexity
Translate network model changes to
OSPF/RSVP representation, handle
concurrent reservations in RSVP
signaling
Same as GMPLS, plus extensions
to PCEP to represent spatial
services
Develop the SDM network model
from scratch, develop or extend
controller and north-bound
interface
Computational
Capability
Typically limited in scope (source
routing based on limited
information) on multiple low power
CPUs
Conceptually encompassing
complex algorithms based on
extensive information and run on
powerful, dedicated hardware
Conceptually encompassing
complex algorithms based on
extensive information and run on
powerful, dedicated hardware
Scalability / Overhead
Slower reactivity due to large
increase in information to
disseminate
Similar to GMPLS: more
computational resource but small
pool of points of failure
Centralized controller gives high
computational resources, south-
bound protocol can limit flooding
Resiliency
High (distributed system), but
partition-crossing services fail
eventually
Only partitions which can reach the
PCEs continue to operate, partition-
crossing services fail eventually
Data plane can use hard
reservations, but CP partitioning
would prevent controlling part of
the network
Programmability
Not supported, and very difficult to
retrofit
PCEP limiting as north-bound
protocol, but could be adapted with
extra software
Supported
Multi-domain/carrier
Supported using e.g. BGP-LS to
flood information, but
confidentiality issues
Supported, if nothing else through
horizontal PCE chains
Open issue
Multi-vendor
Theoretically supported (IETF
standard), but advanced features are
mostly vendor-specific
Theoretically supported, relies on
the underlying GMPLS control
plane
Depends on south-bound protocol,
theoretically supported
INSPACE control plane framework: comparison of
architectural archetypes
IN
Hybrid PCE/GMPLS was the choice for EU projects DICONET & CHRON â INSPACE now shifts to SDN
63. 63
Summary
ď§ The activities of EU project FOX-C were presented
ďŽAll possible spectrally flexible super-channel transceivers were implemented and tested
ďŽNodes for all-optical add/drop of sub-channels (for both NWDM and AO-OFDM multiplexing)
were developed and tested
ďŽNetworking studies were performed to demonstrate the benefits of FOX-C solution
ď§ The activities of EU project INSPACE were presented
ďŽNodes for independent or joint switching of SDM super-channels were developed and tested
ďŽNetworking studies were performed to demonstrate the benefits of INSPACE solution
ďŽDevelopment of an SDN-based control plane for SDM-based networks is underway
ď§ The EU-funded projects FOX-C and INSPACE are developing the
optical networking solutions that will dominate the market after
2020!
⢠Stay tuned!!!
The operating principle of a high resolution switching processor [9] is quite similar to that of wavelength selective switches. However, the bulk diffraction grating dispersive element is replaced by an engineered phase array designed to provide the optical resolution over a finite bandwidth. The spectral switching element is still an LCoS, now operating under much finer spectral granularity.
The proposed architecture is depicted in Fig. 1 a). Two flexible wavelength selective switch units (WSSs), one at the input and one at the output of the node, perform the super-channel selection and re-insertion back to the network traffic. Sub-channel switching is achieved by means of a three branch interferometer structure (branch -A, -B, -C). Each interferometer is capable of single super-channel processing, however, multiple sub-channels can be added or dropped within it at the same time. At the interferometer, a portion of the selected super-channel is dropped to the local receivers and the rest is split between the two branches -A and -B. Sub-channel blocking is facilitated by replicating the corresponding signal waveform and interfering it destructively with the super-channel that propagates at branch-A. This is purely all-optical process taking place in three stages. Initially, the OFDM signal is de-multiplexed by an optical Fast-Fourier Transform processor. Subsequently, a bank of optical gates performs time sampling of the sub-channels. The optical gates are synchronized to a common clock signal, extracted from the super-channel. Fig. 1b) depicts the simulated optical spectrum of an OFDM signal consisting of 7 BPSK modulated sub-carriers at 10 Gbit/s, after the WSS selection. The optical spectrum of the middle (ch 4) sub-channel after the de-multiplexing with an 8-point optical FFT processor [3] is also depicted. The eye diagram of ch 4, see Fig. 1c), has reduced opening due to the preceded matched filtering and the inter-symbol interference (ISI) from the neighboring channels. In this example, crosstalk free regions cannot be identified, as the orthogonality condition has been violated due to the tight pass-band selection of the WSS (3rd-order Gaussian, 100GHz bandwidth) and the finite bandwidth of the transmitter.
The optical gate samples the sub-channels that need to be blocked from the through path (e.g. ch 4). A window of minimum crosstalk, see Fig. 1d) and feeds the sampled waveform to an optical i-FFT processor, with transfer function H(f)=sinc(fT), which reshapes the pulses back to their initial symbol duration T, see Fig. 1e). Finally, the recovered waveform is amplified and interfered destructively with the OFDM signal in the upper branch creating a free spectral position for a new channel to be added, see Fig. 1f). The insertion of new channels takes place on a separate branch (i.e. branch -C), with a bank of laser transmitters aligned to the sub-channel frequencies of the OFDM super-channel. For the alignment known optical carrier extraction and phase locking methods can be applied [4] In this study an ideal phase locking process has been assumed. The resulted optical spectrum, shown in Fig. 1f), and the clear eye diagrams of the added channel (see Fig. 1 h) ) and of its closest neighbors, i.e. ch 5 and ch 3 shown in Fig.1g), Fig. 1i), confirm that successful add-drop operation with low penalty.
More on the channel allocation concept.
- Interfaces have been described in the D5.1
- We plan to develop only one south-bound protocol