Optical Node Architecture and Hitless Spectrum Defragmentation for ...

International Workshop on Optical Networking (iWON’13) Session I: Flex-Grid Networks, Elastic Optical Networks, Optical OFDM Mon, December 9, 2013 9:30-9:50

Optical Node Architecture and Hitless Spectrum Defragmentation for Flexible Grid Optical Networks Xi Wang g(1), Qiong g Zhang g(1), Inwoong g Kim(1), Paparao p Palacharla(1), Motoyoshi Sekiya(1), Kyosuke Sone(2), Yasuhiko Aoki(3) and Hakki C. Cankaya(4) Fujitsu Laboratories of America, Inc., (4) Fujitsu Network Communications, Inc., (2) Fujitsu Laboratories, Ltd., (3) Fujitsu Limited

(1)

Outline of This Talk Fl ibl optical Flexible ti l node d architecture hit t employing software-defined modulation-flexible universal i l ttransceivers i for beyond 100G flexible grid optical networks.

Hitless spectrum defragmentation using g synchronous y bandwidth-variable WSS control for in-service resource optimization.

2

Hakki Cankaya © 2013 Fujitsu Network Communications

Flexible Grid Optical Networks  Beyond 100G 100G, e e.g. g 400G or 1T, 1T will require more than 50GHz spectrum  Super-channel transmission – multiple sub-carriers packed tighter for higher spectral efficiency, managed/operated as single entity  fixed or variable data rate (n x R Gb/s), Gb/s) flexible channel bandwidth (m x slotBW GHz), GHz) design for adaptive reach (modulation format)

 Flexibly adjust spectrum/bandwidth, data rate and reach to match traffic demands – minimize overall resource usage in the network Software-controlled switching between shortand long-haul modes on a single transceiver

16QAM Universal transceiver

Reconfiguration Universal by Software transceiver

Long‐haul Universal Greater than transceiver 1000 km

Short‐haul e 00 A few 100 km Flexible optical switch node

Node1

Node 2

QPSK

Software-controlled adjustment of transmission routes and wavelength slot on an optical switch node

Node 3

Flexible grid optical network 3


Universal Transceiver  Programmable modulation format (DP-QPSK (DP-QPSK, DP-16QAM, DP-16QAM etc.) etc )  DP-QPSK – 100Gb/s  DP-16QAM – 200 Gb/s, same hardware supporting 2x capacity, higher spectral efficiency, lower reach  Reduce sparing for service providers

 Super-channel for 400 Gb/s  Nyquist filtering – spectrum shaping to allow tighter packing of sub-carriers  2x SC DP-16QAM  4x SC DP-QPSK QPSK Tx

Tx DSP

DAC

E/O

Rx DSP

ADC

O/E

Framer

4

Rx

16QAM


Conventional Optical Node Architecture (using dedicated transponders) Client optics

OTU4 framer

Universal transceiver

Type 1: 100G dedicated transponder #1


400G Client optics

OTU framer

100G (DP-QPSK)

Universal transceiver Universal transceiver Universal transceiver

4x 100G (DP-QPSK) (DP QPSK)

#2

Type 2: 400G dedicated transponder 100G DP-QPSK x 4

400G Client optics

OTU framer

#3


2x 200G (DP-16QAM)


Type 3: 400G dedicated transponder 200G DP-16QAM x 2

ROADM

5


Flexible Optical Node Architecture (pool of universal transceivers) Universal transceiver

200G x 2 Client optics

OTU Framer 400G client card

OTU MLD

Universal transceiver Universal transceiver

100G x 4 Client optics

OTU Framer 400G client card

#1 100G (DP-QPSK)

OTU MLD


#2 2x 200G (DP-16QAM)

Universal transceiver Client optics ti

OTU4 Framer F

OTU4/ MLD


100G x 1

100G client card

Universal Transceiver Pool

Cross connect Switch

#3 4x 100G (DP-QPSK)

ROADM

Y. Aoki et.al. “Dynamic and Flexible Photonic Node Architecture with Shared Universal Transceivers Supporting Hitless Defragmentation”, ECOC 2012 6


Performance of Flexible Node Architecture

 Flexible node architecture requires ~20% less transceivers compared to conventional node architecture 7


Spectrum Defragmentation  Signals can occupy different number of slots slots, the uneven usage of slots causes spectrum fragmentation.  e.g. a departed 400 Gbps signal leave 7 empty slots, 4 of which may be reoccupied by a 100 Gbps signal, signal leaving 3 “fragmented” fragmented slots unusable by any signal.

 Hitless Spectrum Defragmentation  The network ‘retunes’ the signals occupying the neighboring slots of a departed signal to fill the λ g p ‘gap’

Signal 1 Slots Vacant

• e.g. the signal occupying the slots above the empty slots always ‘fall down’ to fill the empty slots.

 The transponders and all devices (such as fl ibl grid flexible id ROADM ROADMs)) along l th the signal i l path th ‘retune’ into the new slots.  This defragmentation method supports any topology and works with any routing and spectrum assignment algorithm.

Signal 1

3 Signal g Signal 2

Signal g 4 A

Signal 2

B

C

D

X. Wang et.al. “A Hitless Defragmentation Method for Self-optimizing Flexible Grid Optical Networks”, ECOC 2012 8


Illustration of Spectrum Defragmentation Fragmented spectrum

Optical sspectrum U Usage

Increased usable spectrum as a result of defragmentation

Network et o Topology opo ogy (a) Without defragmentation

(b) With defragmentation 9


Defragmentation Demonstration Using Real-time Coherent Receiver  Four ROADM nodes with flexible grid (A-D) (A-D).  Continuous wavelength sweep-capable tunable LDs (Tx and LO).  Synchronous WSSs operation with wavelength sweep of signal. Defrag Controller

Node IF

WSS

WSS

SMF 80km

15

PN 2 -1

SMF 80km

WSS SMF 80km

Tunable LD (LO)

Node D

E Error Detector

WSS

Node IF

Node C

ADC C/DSP CMO OS LSI

112 Gbps DP-QPSK Transmitter

Node IF

Node B

O O/E Front-end

Tunable LD

Node IF

Node A

WSS Signal λ0

λ1

λ0

λ1

λ0

λ1

R l ti Real-time DSP

K. Sone et.al. “First Demonstration of Hitless Spectrum Defragmentation using Real-time Coherent Receivers in Flexible Grid Optical Networks”, post-deadline, ECOC 2012 10


Experimental Demonstration  Continuous wavelength sweep (over entire C-band) and real-time BER.  Wavelength sweep in steps of 20 pm (2.5 GHz) width every 100 ms. x16 play speed

X-Pol. BER Y-Pol. BER Average BER

Wavelength sweep for spectrum defragmentation without service disruption is successfullyy demonstrated. 11


Key Performance Metrics for Continuous Wavelength Sweep Defragmentation  How H t tune to t the th wavelength l th off tunable t bl lasers? l ?  The faster you can tune, the less amount of wavelength sweeping time.

 How to control the passband of bandwidth bandwidth-variable variable wavelength selective switches (BV-WSS)?  Sequential BV-WSS control or Synchronous BV-WSS control. WSS Signal λ0

λ1

λ0

λ1

λ0

λ1

λ0

λ1

Available Slots λ0

λ1

λ0

λ1

λ0

λ1

λ0

λ1

(a) Sequential BV-WSS control

(b) Synchronous BV-WSS control 12


Comparison of BV-WSS Control Methods  Sequential S ti l BV BV-WSS WSS control t l  The sweeping time is proportional to the product of # of vacated slots and # of signal layers.  It may take quite long time if the departed signal vacates many slots and there are many layers of signals need to be retuned.

 Synchronous BV-WSS BV WSS control  The sweeping time is proportional to the sum of # of vacated slots and # of signal layers.  The overall sweeping time can be significantly reduced reduced. • It also relaxes the hardware requirement on tuning speed of lasers, BV-WSSs, etc., enabling practical and cost-effective defragmentation solutions.

 Synchronous S h BV-WSS BV WSS control t l enables bl more successful f l defrag operations than Sequential BV-WSS control.  Under the same wavelength sweep speed and traffic condition.

13


Blocking Rate Performance

 Reference – 400G over current network using inverse multiplexing (4x100G)  400G Super-channel with adaptive modulation – 85% higher load  Spectrum defragmentation improves network capacity by additional ~15% 14


Blocking Rate vs. Wavelength Sweep Speed Offered load = 100 Erlangs

1.5E-3

Blocking R B Rate

1 3E 3 1.3E-3 1.1E-3

10s Mean Holding Time 100s Mean Holding Time 1000s Mean Holding Time

9 0E-4 9.0E-4 7.0E-4 5.0E-4 5.0E 4 3.0E-4 1.0E-4 1.0E-1

1.0E+0 1.0E+1 1.0E+2 1.0E+3 Wavelength Sweep Speed (ms/2.5GHz step)

 Achieved close close-to-ideal to ideal blocking performance for 1000 s mean holding time with 100 ms/2.5 GHz sweep speed (in our experiment) 15


Summary  Presented P t d a flexible fl ibl optical ti l node d architecture hit t employing l i software-defined modulation-flexible shared universal transceivers to enable beyond 100G transport and in-service resource optimization in flexible grid optical networks.  Spectrum S t d f defragmentation t ti with ith realistic li ti wavelength l th sweep speed of 1 s/2.5 GHz step yields up to 15% higher load than the case without defragmentation for connections with holding times on the order of hours.  The hitless spectrum defragmentation technique can be i l implemented t d iin an existing i ti NMS, NMS or be b iincorporated t d iin a ffuture t Software-Defined Optical Network (SDON) paradigm. y to the deployment p y of p practical and efficient  Pave the way flexible grid optical networks. 16


17
