photonics technologiesphotonics technologies for low
TRANSCRIPT
Photonics technologiesPhotonics technologiesPhotonics technologiesfor low-energy networks at AIST
Photonics technologiesfor low-energy networks at AIST
Shu NamikiNational Institute of Advanced Industrial Science and Technology (AIST)
Network Photonics Research Center
Shu NamikiNational Institute of Advanced Industrial Science and Technology (AIST)
Network Photonics Research CenterNetwork Photonics Research CenterNetwork Photonics Research Center
ISUPT 2013, Oct. 21, 2013At University of Rochester
Part of this work is supported by - Special Coordination Funds for Promoting Science and Technology of Ministry of
1
p g gy yEducation, Culture, Sports, Science and Technology (MEXT).
- New Energy and Industrial Technology Development Organization (NEDO)
TeamTeam• Network Architecture• Network Architecture
– Kiyo Ishii– Junya Kurumida– Tomohiro Kudoh
• Silicon Photonics and Devices– Hitoshi Kawashima– Keijiro Suzuki– G. W. Cong– Ken Tanizawa– Sang-Hun Kim
Haruhiko Kuwatsuka– Haruhiko Kuwatsuka– Hiroyuki Matsuura– Aaron Albores-Mejia
• Transmission and Node SubsystemsTransmission and Node Subsystems– Takayuki Kurosu– Takashi Inoue– Hung Nguyen Tan– Karen Solis-Trapala– Mingyi Gao
2 S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
OutlineOutline
• Introduction:– Scalability, and now is endangered!
• Opportunity for optical switches– Distinct energy scaling and bandwidth– AIST project called “VICTORIES”– Development of 32 x 32 silicon switches
D i O ti l P th N t k• Dynamic Optical Path Network• Inter-Data Center Network• Summary
3 S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
Network traffic keeps on increasingNetwork traffic keeps on increasing
450,000 40% CAGR
But, how sustainable is it for the future?
250 000
300,000
350,000
400,000 AT&T
Verizon
Traffic (A.U.)($m
illion
s)
Gap
40% CAGR
100,000
150,000
200,000
250,000
Rev
enue
Gap
0
50,000
2007 2008 2009 2010 2011 2012 2013
YearYearSource: Fortune 500
4 S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
After T.J.Xia (Verizon) at WIN 2012, Inuyama
Data-centricity of TrafficData-centricity of Traffic
678
arwithin DCD2D (such as VPN)
456
bytes/Yea
Grobal IP DCGrobal IP non‐DC
123
Zettab
01
2011 2012 2013 2014 2015 2016Year
And, most of the contents are video.
http://www.cisco.com/en/US/netsol/ns827/networking_solutions_sub_solution.htmlhttp://www.cisco.com/en/US/netsol/ns1175/networking_solutions_sub_solution.html
S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester5
How have we scaled the transport capacity?How have we scaled the transport capacity?
• 1960s– Copper wire for telephony
• 1970s1970s– Use optical fibers, LDs, PDs
• 1980s– Use ETDM (Sonet/SDH)Use ETDM (Sonet/SDH)
• 1990s– Use EDFAs, Internet revolution!!!
• 2000s• 2000s– Use WDM, Raman amplifiers, FEC, the Bubble!!!
• 2010sUse advanced modulation formats digital coherent technology– Use advanced modulation formats, digital coherent technology
• 2020s– Use Space and/or Mode-Division-Multiplexing
S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
• We’ve been so acclimated to scalable technology, but…6
What we are learning from HPCWhat we are learning from HPC
• Because chips are no longer as scalable, Moore’s law may finally be ending unless the network scales.
A Benner (IBM) OFC 2012A. Benner (IBM), OFC 2012 Tutorial
7 S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
http://www.extremetech.com/computing/165331-intels-former-chief-architect-moores-law-will-be-dead-within-a-decade
Can data centers scale out?Can data centers scale out?
• The role of network gets more important.
8 S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
OIDA Workshop Report April 2013
Fundamental limits being faced!Fundamental limits being faced!• Quantum Limit• Quantum Limit
– Distributed Raman amplifiers with 3.1 dB noise figure– E-FEC allows error free transmission with only several photons per bit
at receiverat receiver• Rayleigh Scattering Limit of Fiber
– ~0.15 dB/kmShannon Limit• Shannon Limit– A few dB away from E-FEC technologies– Non-linear Shannon limit is about 100 Tbps/fiber
El i l B d id h Li i Th Li i f P ll li• Electrical Bandwidth Limits = The Limit of Parallelism– Radiation loss– Thermal jitter– Bandwidth density = energy consumption / bit
• TCP/IP Limits– Increasing inefficiency for larger Delay-Bandwidth products– QoS limitations– IP routers consumes energy proportional to the amount of traffic
9 S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
Energy issue will eventually take over everythingEnergy issue will eventually take over everything
• Green of ICT, then Green by ICT 1,000
100,000
10,000Japan’s Case
100,000
tion
of IP
Rou
ters
1,000
10,000
(Tb
ps)
• Disaster proof10
1
100Mostly High-
Definition Video Transmissions
l ene
rgy
cons
umpt
(TW
h)
1.0
10
100
tal I
nter
net T
raffi
c Japan’s TotalElectricity Generation
In 2005
p– Communications is the
‘Lifeline’.2000 2010 205020402020 2030
0.1
• The current technologies can’t scale to the increasing traffic in future.• 3-4 digit energy saving is necessary, which means we need a new paradigm.
Ann
ual
0.1 To
– Energy supply will be limited when disasters occur.
10 S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
Almost all of the traffic will be due to video transferAlmost all of the traffic will be due to video transfer
Consumer Internet Traffic1.5
Traffic Growth Projected by CISCOTraffic Growth Projected by CISCO
Web, Email, and DataFile Sharing(P2P)VideoGamingVoice/M
onth
)
Percentage from 2008 to 2013
ffic
(EB
/
VideoCAGR=1.70 CAGR=1.39
1.022.8%→61.6%
from 2008 to 2013
0.5
net T
raf
File ExchangeCAGR=1.21 50 8%→24 7%
After MIC(0.32 EB)
Video
0
Inte
r
14.9%→8.2%Web, CAGR=1.24
7 8%→1 8%3.7%→3.7%
50.8%→24.7%
02008 2009 2010 2011 2012 2013
7.8%→1.8%
source: Cisco Visual Networking Index: Forecast and Methodology, 2008-2013
11 S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
Packet switching versus circuit switchingPacket switching versus circuit switching
IP: Process every packet for routingy p g
Which suits better for video?
12 S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
Optical Path: Set up end-to-end paths
Switches: photonics vs. electronicsSwitches: photonics vs. electronicsO ti l it h El t i l S it hOptical switches Electrical Switches
Type of switching Circuit switching(requires a control plane)
Packet switching
Granularity Coarse and inflexible Fine and flexibleBandwidth scaling Transparent Transceivers and LSIPhysical I/F Optical connectors TransceiversEnergy scaling < W/port (incl. amp) 0.3 nJ/bit (Infiniband)
6 nJ/bit (IP router)Energy consumption < 1 kW 300 kW (Infiniband)gy pfor 1,000 ports x 1Tbps
( )6,000 kW (IP router)
Scalability issue Number of ports Moore’s lawQoS Physically guaranteed DependsQoS Physically guaranteed DependsSuitable apps Video Packetized data
Optical switch can save energy by several orders of magnitude!!
13 S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
Optical switch can save energy by several orders of magnitude!!The hybrid use of packet and optical switching will be the key.
Difference at the network level..Difference at the network level..C ti ll h di it l k t A t d T t• Conventionally exchange digital packets: Aggregate and Transport
10 100 GbE 100-1000 GbEA few lines between Routers
Bottle10-100 GbE
Statistical
100 1000 GbEBottle-neck Bottle-
neck
A N S h E h d t d fib i ll l!!
Statistical Multiplexing
Statistical Multiplexing
Many CRS-3 required!!
• → A New Scheme: Exchange end-to-end numerous fibers in parallel!!・ SDM rather than TDM・ Full mesh connections
rather than
N usersTransmission lines in parallel space
rather than Aggregate and transport
S Namiki et al ECOC 2010 Mo 2 A 4N x m
m paths
14 S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
S. Namiki et al., ECOC 2010, Mo.2.A.4.S. Namiki et al., JSTQE, 17, 446 (2011).
N x m Matrix Switch Optical switches:
Transparent and Green
It calls for new network technologiesIt calls for new network technologies
Energy and CostEfficiency
AgilityResilience
L3 (IP) Routing
L2 (Ethernet) Switching Poor
Good
L2 (Ethernet) Switching
L1 (ODU) SwitchingDynamic
Wavelength Path Switching
Poor
Good Control Plane at
upper layers
Dynamic
Optical PathWave Band Path Switching
Optical Path Switching
GoodPoorpp y
&Good Optical
Devices
p
Network
15 S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
Optical Path Switching Devices(DOPN)
Toward Optical Cut-Through of Everything (OCE)Toward Optical Cut-Through of Everything (OCE)
• TodayAggrAggr
CoreCoreEdge Edge
Aggr.Aggr. Aggr.
Aggr.AccessAccess AccessAccess
OLT OLT
pSplitterpSplitter
ONUONU ONUONU
16 S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
Toward Optical Cut-Through of Everything (OCE)Toward Optical Cut-Through of Everything (OCE)
• TomorrowAggrAggr
CoreCoreEdge Edge
Aggr.Aggr. Aggr.
Aggr.AccessAccess AccessAccess
OLT OLT
pSplitterpSplitter
ONUONU ONUONUOptical Cut-through of Core Routers
17 S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
Toward Optical Cut-Through of Everything (OCE)Toward Optical Cut-Through of Everything (OCE)
• Ideally Energy-Saving DOPNAggrAggr
CoreCoreEdge Edge
Aggr.Aggr. Aggr.
Aggr.AccessAccess AccessAccess
pSplitterpSplitter
OLTDOPN-FE
OLTDOPN-FE
ONUONU ONUONUOptical Cut-through of Core + Edge Routers
Global Control Plane and Resource Management are necessary
18 S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
Global Control Plane and Resource Management are necessary.
Jump from Optical Cut-through to DOPNJump from Optical Cut-through to DOPNModel for Dynamic Optical Path Network (DOPN)Core
IE
Big Datacenter
Model for Dynamic Optical Path Network (DOPN)
ODU-SWL2
Transponder
WSS
制御I/F
Core
Data Center
Core router15.4nJ/bit
Edge router
Transponder42W/10Gbps149W/40Gbps351W/100Gbps(DC)
Optical amplifier138W/fiber/60km
REG
Regenerator(comparable energy
consumption with TPND)
Link distances: 150~850km
・Avg. hop number is 4.2.
Logically, 4x4 regular lattice in corewith multi-granular switching.
IESmall DC orContent server
SW
Metro-core
Aggregation
OLTOLT Access
Edge router19.0nJ/bit
Edge switch 30nJ/bit**Considering low utilization of down-link side ports
Optical Node86W/degree
400W/equipmentA d ti 410 ONU
ONU
Splitter
Intelligent edge (IE)
OLT ODU Agg.OpenFlowSW
ONU ONU ONU
Splitter
IE IEPath requests
OLT OLT
Splitter
ONU ONU
Splitter
ONUONU
OLTOLT
Splitter
ONU ONU
Splitter
ONUONUTraffic distribution defined as a function
of population distribution
AccessAccommodating 410 ONUs(80% utilization)
5W/equipment
Calculation by K. Ishii (AIST),Under the NEDO Green IT project.
Calculation by K. Ishii (AIST),Under the VICTOIRES project.Details in Photonics West 2013.
100
1000
on [TWh/year]
Extrapolation of the present modelOptical cut‐through in coreDynamic optical path network
Yr 2030150
200
250
300
tion [TWh/year]
Level 1
Level 2
Aggregation
OLT
ONU
Optical cut-through in core
jDetails to appear in ECOC2013.
1
10En
ergy Con
sumptio Yr. 2030
Yr. 2024
0
50
100
150
2008 2010 2012 2014 2016 2018 2020 2022 2024 2026 2028 2030
Energy Con
sumpt
Extrapolation of the present model
19 S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
10 10 20 30 40
Total Traffic [Pbps]
2008 2010 2012 2014 2016 2018 2020 2022 2024 2026 2028 2030Year
Traffic CAGR: 40%The energy reduction by optical cut-through is not large enough.The energy consumption at edge routers is dominant.
Some initiatives have already startedSome initiatives have already started
• GreenTouch Consortium • VICTORIES
http://www.greentouch.org/
IEC’s Megatrend: Energy Sustainability of optical communication networks
20 S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
http://www.aist-victories.org/
IEC s Megatrend: Energy Sustainability of optical communication networksand their driver: Dynamic Optical Path Network
Digital Optical Path Processor using Silicon PhotonicsDigital Optical Path Processor using Silicon Photonics
32 32 M t i S it h• 32x32 Matrix Switch– Energy: Need only a few Ws– Footprint: Within 2 cm x 1 cm Chip
Switching time• Switching time– ~ sec by thermo-optical switching – ~ nsec by plasma-effect switching
• Integration with control circuitsCourtesy of Y. Shoji, K. Kintaka, et al.
• Integration with control circuits for black-box operations
TO-SwitchOpitcal Fiber Array
Control Circuit
p y
S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester21
Why 32 x 32 matrix switches?Why 32 x 32 matrix switches?
• 32 x 32 SWs make a 512 x 512 Clos SW.• 100,000 CPU nodes can be accommodated by 675 450x450
(simplex) switches A 450x450 SW consists of 96 32x32 SWs(simplex) switches. A 450x450 SW consists of 96 32x32 SWs, leading to a total of 64,800 32x32 SWs.
• Each port can carry a bandwidth > Tbps.
22--stage Fat tree Switch Fabric for 100,000 nodesstage Fat tree Switch Fabric for 100,000 nodes
450x450 450x450450x450 225 switches450(d)
450x450 450x450 450x450 450x450 450 switches
duplex 225 portssimplex 450 ports
Total: 101250 CPU nodes
S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester22
Roadmap of Si-photonics matrix switch at AISTRoadmap of Si-photonics matrix switch at AIST2014
32×322014
2013
Mach-Zehnder Interferometer based TO switching
2012PILOSS8 x 8
2013
PILOSS4×4
2011~20 mm
4×42010
20 mm
2×2
23 S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
2×2 Polarization diversity2×2 Polarization diversity
PS PS2X2 Sw.Unit for TE
TE
Kim et al., in ECOC2013
PS PS
Intersection
2X2 Sw.Unit for TM
Orthogonal SOP Crossing(JP 2011-158030 )
TM
(JP 2011 158030 )
TETM
25 S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
Si-ph. SW integrated with CMOS driver circuitsSi-ph. SW integrated with CMOS driver circuits
• OXC: 2x2 TO switch with MOSFET– G. W. Cong et al., OE March 2013 (AIST)
• Interconnect SW: 4x4/8x8 p-i-n switch with CMOS driver– B.G. Lee et al., OFC NFOEC PDP 5C3, 2013 (IBM)
26 S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
Realizing 32 x 32 chip scaleRealizing 32 x 32 chip scale
20 mm×10 mm
Drawing time using Spot-Beam EB Lithography System (JEOL6000FS/E (50kV)) 32×32
8×811 mm×7 mm
8×8128 hrs. (~5 days) by EB
6 mm×3 mm4×4
8 hrs.
DUV photolithographyEB
27 S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
2 hrs. ArF Immersion Lithography facilitySuper Clean Room (SCR) 12 inch line @AIST
Hierarchical multi-granularity in routing
• Various path granularities– From fine to coarse - node that can handle such
Fiber
1GE
10GE
Sub-lambda path
10GE pathOTNODU
ODU
ODU
100Gbit/s~
制御I/F
光パスFiber path
Control I/F
path
pathOTN
ODU-SWL2
Transponder
WSS 10~100 Gbit/s
電気スイッチ
光スイッチ
波長パスLambda pathOptical SW
Electrical SW
ODUSW
Lambda SW(WSS)
Fiber path SW(Matrix SW)
SW
1~10 Gbit/s サブ・ウェーブレングス・パスSub-lambda path~1Gbit/s
MPLS-TP
• Network topology based on hierarchical routingL ti S l bilit t ti l l
K. Ishii et al., Photonics West 2013
28
– Low energy consumption -Scalability to national scale
S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
CDC-ROADM using Silicon Photonics (NEC+AIST)
次世代ROADM1:8 Splitter 8x1 Switches
λ1~λ8
BER
10-5
10-6
10-7
10-8
… …
WSS/Splitter
WSS/Splitter
Next Gen. ROADM
λ1~λ8
In-Port 1 In-Port 8
λ1 λ843.018 Gbps DPSK signalOSNR 21 dB @ res 2 nmReceived Power: ‐9dBm
Out-Port 3 Out-Port 4 Out-Port 5
Out-Port 6 Out-Port 7
10-9
10-10
10-11
48 transponders
…..
48 transponders
…..
/Splitter /Splitter
1 x 8 CPL 1 x 8 CPL
Drop AddSi-TPA transmission card
… … … … … …
16 mm1:8 Splitter 8x1 Switches
48 transponders 48 transponders
8 Input/outputfibers
Transponder
TPAmodule
TPAmodule
TPAmodule
TPAmodule
TPAmodule
TPAmodule
.. .. ..
… … … … … …
.. .. ..
SWDrivers
12 mm
Gate / SelectorSplitterTo input/output fibers
(8-port)
To transponders(8-port)
152 switch elementsintegrated
FPGASwitch drivers
48 transponders
TPA module x3 / daughter card(8 x 8 Si switch module)
TPA module x12
T. Hino et al., ECOC2012 Invited:Tu.3.A.5
29 S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
Feature of ODU XC (Fujitsu+AIST)Feature of ODU XC (Fujitsu+AIST)
• ODU XC (Cross Connect) is a digital cross connect for ODU (Optical Data Unit) that defined by ITU-T G.709.
ODU XC has capability of switching traffic with
OpticalFibre
OTN
OTNODUODUODU
1GbE1GbE10GbE
λPath
λ• ODU XC has capability of switching traffic with 1.25Gbps granularity. It is very efficient to switch 1Gbps-40Gbps stream.
• ODU XC has 1/3 power consumption and 1/100
OTN λPath
ODU XCODU XC has 1/3 power consumption and 1/100 latency compared to core router.
• AIST and Fujitsu are developing the ODU XC chip and this chip will be available in 2013 .
ODUk(k=0,1,2,3,4,flex)ODUk(k=0,1,2,3,4,flex)ODUk(k=0,1,2,3,4,flex)ODUk(k=0,1,2,3,4,flex)
…ODUk(k=0,1,2,3,4,flex)ODUk(k=0,1,2,3,4,flex)ODUk(k=0,1,2,3,4,flex)ODUk(k=0,1,2,3,4,flex)
…
p
λ 40GEODUODUXCXCODUflex#1Packet#1
ODU210GE ODUODUXCXCR t
Packet#1ODUflex#1Packet#3ODUflex#3 Router
OTN is effective and efficient f G
λλ
λXC XC nodenodeODUflex#2
ODUflex#1Packet#1Packet#2100GE
XC XC nodenode
Router Packet#3ODUflex#3 Routerfor operating >1 GbE. Fujitsu plans to incorporate the OTN switch fabric into Fujitsu products.
S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
λODUODUXC XC
nodenode
ODUODUXC XC nodenode
ODU2ODUflex#2
10GERouterODUflex#3
Packet#240GE
Packet#310GERouter
products.
30
Slide courtesy of H. Honmaand H. Onaka, Fujitsu
32 x 32 SWs make up a nationwide Dynamic Optical Path Network
32 x 32 SWs make up a nationwide Dynamic Optical Path Network
R l IP t b ti l it h !!
Wavelength path terminal
Sub- aggregating node
Sub- and wavelength cross-connect node
Hierarchical multi-granular node
Replace IP routers by optical switches!!
WSS
100Gbit/s~
10~100 Gbit/s
C-plane I/F
Fiber path
-path
………
…………n
Sub- path terminal
Core
Multi-granular aggregating node
Content server
ODU-SWL2
Transponder
WSS
1~10 Gbit/s
Electrical
Optical
Sub- path1 2 3 …
…………
1 k
j11j
Category-core
K. Ishii, et al.(AIST), Photonics West 2013j
1
jCategory 12
3 … ……k
12
3 … ……k
11
23 … …
…k
12
3 … ……k
1 mx mmx
… … … …
1 mx+1
m
1 2 i
mx
1 2 i
1 mx+1
m
1 2 i
mx
1 2 i
1 mx+1
m
1 2 i
mx
1 2 iGroup
PassiveDouble Star
OLT
31S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
Regular mesh in Core; Full Mesh in Metro; Complete bipartite graph in AccessFlex-Grid compatible, ODU-XC for sub- granularity
Inter-datacenter network: challengesInter-datacenter network: challengesCh k f ti ll t d d t• Chunk of continually generated data
• Ultra-coarse granularity• Low latency data transfer• High capacity and efficiency• Strict QoS with scheduled connections
Datacenters
DistributedDistributedDatacenters
• Seamless connections to intra-datacenter networkSeamless connections to intra datacenter network• Resilience for disaster
32 S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
Multi-Tbps Elastic Transmitter : WDM vs. OTDMMulti-Tbps Elastic Transmitter : WDM vs. OTDM
S h l N i /OFDM WDM• Super-channel Nyquist/OFDM WDMMod.xN carriers
・・・
CombGen.
NxmMatrix
SW
Mod.
Combiner・・・
AWG
PDM
• Nyquist OTDM WDM
Mod.Combiner
DSP hungry
• Nyquist-OTDM WDM0xMod.xN symbol slots
CombGen.
1x
Splitter Combiner・・・
Mod.
M d
・・・
Wave-shaper+ PDM
33 S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
mxMod. Optical processing needed?
Optical comb generator (OCG) for Tbps elastic Tx/RxOptical comb generator (OCG) for Tbps elastic Tx/Rx
• High spectral efficiency through superchannel (OFDM or Nyquist)• High quality optical comb/ultra-short pulse through highly nonlinear fibers
– V. Ataite et al. (UCSD), OFC NFOEC 2013, PDP5C.1V. Ataite et al. (UCSD), OFC NFOEC 2013, PDP5C.1
C+L-band40 dB OSNR
– T Inoue and S Namiki Laser & Photon Rev 2: 83–99 (2008)T. Inoue and S. Namiki, Laser & Photon. Rev., 2: 83 99 (2008).
< 400 fsecTunableTunable
• R&D on Tbps elastic transceivers using OCG is highly expected.
34 S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
Research on Nyquist OTDM‐WDM signalsResearch on Nyquist OTDM‐WDM signals
oHigh spectral efficiency up to 1 symbol/s/HzoUltrahigh yet flexible baudrates (up to 1T)
Features of Nyquist OTDMFeatures of Nyquist OTDM
oGeneration in optics, high energy efficiencyo Eliminating guard‐bands in WDM systemo Better transmission perf. than OTDM
T Hi k t l O t E 20 2012
M. Nakazawa et al., Opt. Express, 20, 2012.
o T. Hirooka et al., Opt. Express, 20, 2012.o K. Harako et al., OFC 2013, JW2A.38.o H. Hu et. al., CLEO 2013, CTh5D.5.
G. Bosco et al., J. Lightw. Technol., 29, 2011.
i i ki i koGranularity will grow to be more than ~100G in the future big‐data centric era
o Especially traffic between data‐centers
Nyquist OTDM‐WDM in networkNyquist OTDM‐WDM in network
o Especially, traffic between data‐centers
oHighly efficient, ultra‐coarse yet flexible granular inter‐datacenter network based on
35 O. Gerstel, et al., Comm. Magazine, 50, 2012.
all‐optical elastic network Nyquist OTDM‐WDM signals
S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
Nyquist filtering using WaveshaperNyquist filtering using Waveshaper
Ultra‐coarse granularNyquist signalsyq g
Ultra coarse granular Nyquist signals (>40G) can be generated by WaveshaperoUltra‐coarse granular Nyquist signals (>40G) can be generated by WaveshaperoWaveshaper‐based WSS can efficiently add‐drop WDM channels of Nyquist OTDMs with almost no guard‐band
o Pass drop over cascaded WSSs: H Nguyen Tan et al OFC 2013 JW2A 50
36
o Pass‐drop over cascaded WSSs: H. Nguyen Tan et al., OFC 2013, JW2A.50 This paper: transmission and pass‐drop operations in full elasticity
S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
Experimental setupExperimental setup
oBaudrate/pol. : 43Gbaud, 86Gbaud, 172Gbaud, and 344GbaudoClock recovery by optical null‐header (ONH): Kurosu et al., Opt. Express, 21, 2013.oTransmission 4 spans of 80km SLA+IDF and WSS nodes
37
oTransmission 4 spans of 80km SLA+IDF and WSS nodesoWSSs elastically pass‐drop WDM channels of Nyquist OTDMs
S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
Optical spectra of Nyquist OTDM WDMOptical spectra of Nyquist OTDM WDMH. Nguyen Tan et al., OECC 2013, PDP / ECOC We.1.C.5Details to appear in OE.
WSS nodeINPUT
PASS
DROP
o Simple generation of different baudrate signals by Waveshaper
oWSS efficiently add‐drop WDM channels of Nyquist OTDMs with almost no guard‐band
38 S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
BER performance of Nyquist OTDMsBER performance of Nyquist OTDMs
ONH
ONH
ONH
39 S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
VICTORIES Test Bed DEMO2014VICTORIES Test Bed DEMO2014O t 2014 i T k b J• Oct. 2014 in Tsukuba, Japan
• Co-located with a 3-day international symposium• Demonstrate 4k/8k video apps over DOPN• Demonstrate 4k/8k video apps over DOPN
N2:ダイナミック波長パス
N1:多粒度ダイナミック・ノード(AIST)
P1, P2:シリコンフォトニクスTOスイッチ(AIST)(富士通研)
ネットワークリソースマネージメント I1:ストレージ資源管理(AIST) I2:ストレージ・ネットワーク統合資源管理(NTT)
P5:集積デバイス実装(古河電工)
TL‐1
ITRI M i NPRC EAST1Di ti 3 Direction 2
TL‐1 TL‐1
N2:ダイナミック波長パス・ノード技術(NEC)
N3:サブウェーブレングス・パス・ノード(富士通)
N4:ダイナミック波長可変光源(住友電工)
P3:トラポン・アグリゲータ(NEC)
配信サーバB(10G, HD)視聴者SA (8K)
コンテンツ集積拠点(HD)
Mon.
P4:次世代WSS(日立電線)ODU SW
TransponderAgg.
ODU SW
TransponderAgg.
(古河電工)
TV会議SYS2
配信サーバC(10G, HD)
視聴者B視聴者A
C5:超高速波形観測技術(Alnair) C4:集積型分散モニタ技術(フジクラ)
ITRI‐Main NPRC‐Main
ITRI‐Main‐TKB‐WEST(NS1)
配信サーバA(1G, HD)
Agg.
NPRC‐EAST1Direction: 3 Direction: 2Direction: 3
ODU SW
C1:自律制御光パス・コンディショナ開発(AIST)
C2:高速制御技術(Trimatiz)
C3:特殊ファイバ、高密度線路技術(古河電工)
拠点(HD)サーバS
MUX
Agg.
視聴者CC6:光パス多重技術(富士通研)
Agg.
AIST情報棟
AIST デモ会場視聴者D
ITRI‐WEST NPRC‐CTRDirection: 3
Direction: 4Agg.
視聴者E
NPRC‐EAST2
視聴者SB (8K)
Direction: 3
制御 素 確 波 帯域 ポ
Agg.
TV会議SYS1 視聴者Fつくば地区
デ 会場視聴者D
TKB‐SUB2
視聴者SC (8K)
Agg.
TKB‐SHV Direction: 2
Direction: 2
TKB‐SUB1Direction: 3
40 S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester
配信サーバ(8K, SHV)
・制御要素の明確化(波長,帯域,ポート#,フォーマット,ビットレート,ファイバ種,分散量) 協力機関
SummarySummary• The energy consumption of the IP network is becoming a serious bottleneck,
calling for a paradigm shift.• For disaster survivability, resilience under limited energy supply is of critical
importanceimportance.• Dynamic Optical Path Network may be the only option in the sense of both
energy, capacity, and suitability for video services, thus has been recognized by IEC as a driver of this megatrend.y g
• In order to realize this, some R&D initiatives have been put into action.• We have reviewed some of the activities under VICTORIES project at AIST.• An extensive effort on developing 32 x 32 silicon photonics switches wasAn extensive effort on developing 32 x 32 silicon photonics switches was
reviewed.• All optical Nyquist filtering works well higher bandwidths than 40 Gbaud
signals.
• Jump from ongoing optical cut-through to Dynamic Optical Path Network must be assisted through academic activities and standardization!
• Orchestration of different layers and/or standardization bodies is essential.
41 S. Namiki, ISUPT2013, Oct. 27, Univ. of Rochester