photonics enabling the zettabyte network evolution1122 fault localization in an optical network...

loukas 1

1

Photonics Enabling the Zettabyte Network Evolution

Loukas Paraschiscisco

2

Future Household Bandwidth Requirements

Twenty future homes would generate more traffic than the entire 1995 Internet backbone.

Cisco estimation: ~ TB/month == HDTV + SDTV + PVR + VoIP + HSD

Household Bandwidth Needs in 2010 (U.S.)

© Loukas Paraschis, cisco, 2009. All rights reserved.

loukas 2

3

Emerging Zettabyte IP Networks

http://www.cisco.com/en/US/netsol/ns827/networking_solutions_sub_solution.html

Perspective: 10 Exabyte = 50x world print (or 2x words ever spoken) Perspective: 10 Exabyte = 50x world print (or 2x words ever spoken) © Loukas Paraschis, cisco, 2009. All rights reserved. © Loukas Paraschis, cisco, 2009. All rights reserved.

44

Operationally enhanced evolution is very important in network innovation…


loukas 3

5

Outline


� Traffic Evolution � WDM Evolution

� ROADM and WXC Networking� 40 and 100 Gb/s Transmission

� Network Architecture Evolution � Packet & WDM Transport Convergence r&D� Flexible/Adaptive Transport R&d

� Summary© Loukas Paraschis, cisco, 2009. All rights reserved. © Loukas Paraschis, cisco, 2009. All rights reserved.

6

Acknowledgements

� Many colleagues at cisco, especially in the Emerging Markets, Core Routing, Optical, and URP, and particularly Dr. Ori Gerstel

� interactions (around the world) with service providers, network equipment, and academia, particularly A. Willner (USC), and B. Yoo (UC-Davis).

© Loukas Paraschis, cisco, 2009. All rights reserved. © Loukas Paraschis, cisco, 2009. All rights reserved. © Loukas Paraschis, cisco, 2009. All rights reserved. © Loukas Paraschis, cisco, 2009. All rights reserved.

loukas 4

© Cisco Systems, 2006. All rights reserved. [email protected]© Cisco Systems, 2006. All rights reserved. [email protected] 777777© Loukas Paraschis, cisco, 2009. All rights reserved. © Loukas Paraschis, cisco, 2009. All rights reserved.

* T. Li and Herwig Kogelnik * T. Li and Herwig Kogelnik

11stst Gen. OFC = TransmissionGen. OFC = TransmissionBi

t Rate

Bi

t Rate

--Dist

ance

( Gb/s

Di

stanc

e ( G

b/s ��k

m)km)

19701970 1975 1980 1985 1990 1995 2000 2005 1975 1980 1985 1990 1995 2000 2005 YearYear

11101011

101022101033101044

101055101066101077

��

��

��

��

��

��

��

��

WHAT’S NEXTWHAT’S NEXT ????�� WDM + WDM + Optical AmplifiersOptical Amplifiers�� Optical AmplifiersOptical Amplifiers�� Coherent DetectionCoherent Detection�� 1.51.5µµmm SingleSingle--Frequency LaserFrequency Laser�� 1.31.3µµmm SM FiberSM Fiber�� 0.80.8µµmm MM FiberMM Fiber

��

Line AmplifiersLine AmplifiersMMUUXX

DDEEMMUUXX

© Cisco Systems, 2006. All rights reserved. [email protected]© Cisco Systems, 2006. All rights reserved. [email protected] 888888

IP/MPLS

ATM / Ethernet

SDH/OTN

DWDM

Multi-service Network Architecture

loukas 5


RegionalRegionalHubHubMetroMetroAccessAccess

CustomerCustomerLocation(s) Location(s)

2nd Gen OFC: Multi-service WDM OADM Transport

Increase Revenues (New Services)Increase Revenues (New Services)Increase Revenues (New Services)Increase Revenues (New Services)

Exist

ing TD

M PL

Exist

ing TD

M PL

10/10

0 PL

10/10

0 PL

Line R

ate GE

Line R

ate GE

Video

Video

SAN

SAN

Multi

Multi--

point

point

Multi

Multi--

point

VPN

point

VPN

Wavelen

gthWa

velen

gth

Lowe

r Cos

t/Bit

Lowe

r Cos

t/Bit

(Dec

rease

d Netw

orkin

g Cos

ts)(D

ecrea

sed N

etwor

king C

osts)

•• Flexible transport (RPR/SONET/Flexible transport (RPR/SONET/λλ))•• Automation (ROADM, software)Automation (ROADM, software)•• Pluggable opticsPluggable optics


Flexible & Scalable Multiservice WDM Transport Challenge: Common low-cost transmission engineering for different applications • Add/drop percentage and location flexibility • Multiple service rates and interface types• Channel number variation (fiber cuts/traffic changes)• Long-term aging effects • Minimize regeneration

Solution: Advanced System Design• Automatic power control• Advanced EDFA (variable-gain, and transient suppression) gain tilt control, ASE suppression• Low-loss OADM, DCF (>250 FoM),…• Forward error correction• Metro-optimized transponders

0

5

10

15

20

25

30

14 16 18 20 22 24 26 28Nominal Optical Amplifier Gain [dB]

Span

Loss

[dB]

Number of Fiber Spans1

23

45

11

1010

100100

10001000

Total

capa

city (

Gb/s)

A - Z distance (km)11 1010 100100 10001000

Metro access

Metro/Regional//Long Haul

Extended/UltraLong Haul

MSTPMSTP

loukas 6


AutoPowerControl for Multi-service WDM Transport

90.1Km 22.5dB

32.3Km 8dB 27.3Km

6.5dB 72Km 18dB

28.5Km 7dB

4dB 10.7Km 2.8dB

10dB

5dB

5dB

10dB 10dB

3dB

5dB

5dB

10dB

FOADM #1

BOADM #9

BOADM BOADM

#11 FOADM

#3 BOADM

#12

BOADM BOADM

BOADM

BOADM

BOADM

BOADM #3

BOADM OLA

BOADM

FOADM#2

Goal #1: Maximize OSNR•minimize EDFA ASE, gain tilt, • power variation (max channel min ONSR), Goal #2: Operational SimplicityInstallation, Operation, Maintenance

Solution: Monitor & Control power, coordinating around ring � Installation: pre-provisioning/optimization, � Operation: const Gain sub-ms transients (local) � Maintenance: Pout account aging (OSC/EMS) � Intelligence: like the ASE correction (“true channel power”) allows 10G-FEC spec 7x20dB !

24dB OSNR at 10Gb/s in 1524dB OSNR at 10Gb/s in 15--node 120dB channels node 120dB channels

Paraschis et. al OFC 2005Paraschis et. al OFC 2005

1212

Fault Localization in An Optical Network� Optical networks were harder to troubleshoot

due to their analog nature� Historically:

� No signaling to correlate alarms� Not enough visibility into the signal� Can’t look into the bits at middle nodes

� Solved by MSTP:� Implements fault localization� Integrated photodiodes everywhere for

max visibility

Embedded signaling

1011010…1011010…1011010…1011010…

© Loukas Paraschis, cisco, 2009. All rights reserved. © Loukas Paraschis, cisco, 2009. All rights reserved.

loukas 7

1313

�� Transparent Open Transmission Transparent Open Transmission EFEC, adv. mod., EFEC, adv. mod.,

�� Operationally Friendly Operationally Friendly G.709 OAM&P, tunability, monitoringG.709 OAM&P, tunability, monitoring

�� Network planning flexibility Network planning flexibility ROADM, Planning tools ROADM, Planning tools

3rd Gen: Reconfigurable Open WDM Transport

$0

$1,000

$2,000

$3,000

$4,000

$5,000

$6,000

$7,000

$8,000

$9,000

Reve

nue (

$M)

CY02 CY03 CY04 CY05 CY06 CY07 CY08 CY09

Year

Worldwide M etro Optical Network Hardware M anufacturer Revenue Breakdown

WDM Switch (ROADM)

WDM Transport

SONET/SDH Switch

SONET/SDH Transport

-3

-2

-1

0

1

2

3

4

1525 1530 1535 1540 1545 1550 1555 1560 1565

Wavelength (nm)

(dBm

)Amplitude (w/o WSS)

Amplitude (w WSS)MultiMulti--degree degree ROADMROADMROADMROADM ROADMROADM

Paraschis et. al OFC 2005


14

New DWDM Operational Paradigm

Traditional DWDMTraditional DWDM� Fixed Filters� Fixed Transmit Lasers� Manual or Semi-automatic VOAs

� Manually Configured Amplifier Gain

� Manual Configuration of DWDM Parameters

NextNext--Gen DWDMGen DWDM� Reconfigurable Filters � Tunable Lasers� Automatic VOAs� Automatically Configured Amplifier Gain

� Design Tool Provisions DWDM Parameters


loukas 8

15

Optical Innovation

� Flexible/Reconfigurable OADM, � Optical Module integration; � Tunable lasers � Enhanced FEC, � Electronic (post-Detection)

Equalization, � Advanced Modulation, � Advanced amplification, � Dispersion compensation, � Performance monitoring, � 40/100G upgrade improve network deployment cost

(CapEx), density, and flexibility (OpEx).


loglog(BER

)(BER

)

44 55 66 77 88 99 1010 1111 1212 1313 1414 1515--1515--1414--1313--1212--1111--1010--99--88--77--66--55--44--33--22--1100

SS /N/N [dB][dB]

UncodedUncodedG.709G.709RS(255,239)RS(255,239)

raw channel BER=1.5eraw channel BER=1.5e--33

NECG=8.4 dBNECG=8.4 dBNECG=6.2 dBNECG=6.2 dB

16

Optical Tx/Rx Module Designs

300 Pin MSA300 Pin MSA

300 Pin Small Form 300 Pin Small Form Factor MSA Factor MSA

200 Pin MSA200 Pin MSA

XFPXFP

0.040.04

0.150.15

0.200.20

0.130.13

1.461.46

XENPAKXENPAK

XPAKXPAK0.260.26


loukas 9

17

Optics energy and density Innovation� J. Eng, Finisar, 2009 OIDA AF

18

Outline






loukas 10

19

WDM Transmission Evolution to 40 and 100 G

• Higher rate initially deployed in highly congested links lower TCO vs higher $/bit/s/km/channel

• Higher rate channels (= less wavelengths) preferred (less HW & managements)

• Higher rate preferable over IP link bundling

• Mainstream deployments require operational parity (OSNR, PMD), TCO advantage


20

Advanced Modulation for 40/100 Gb/s WDM evolution


loukas 11

21

40G WDM Adoption – in progress…

170,768158,118146,406127,310

102,66974,39859,04647,23738,09431,483

111,833100,75190,76680,32469,24553,67844,73235,78630,07227,092 87,348

62,39143,93830,512

20,89914,02610,1649,49910,10510,637 85,43736,35616,1587,3451,6328845000

3,20269696 050,000100,000150,000200,000

Port

Shipm

ents

CY02 CY03 CY04 CY05 CY06 CY07 CY08 CY09 CY10 CY11100G

OC-768

OC-192OC-48

Below OC-48

Calendar Year

Worldwide WDM Port Shipments by Speed

Infonetics May 08

Year 1999 1999 2006 2006 2006 2008 2011 2011Gb/s 2.5 10 2.5 10 40 40 40 100

OSNR 11 17 5 6 13 7-8 6 < 10CD 2000 400 5400 1200 150 700 (TDC) 1000 > 400Density 2 4 0.2 (sfp) 1 2 2 1 1-2CostCost 66 2020 ~ 1~ 1 44--55 > 25> 25 TBDTBD < 10< 10 TBDTBD


22

100G WDM Adoption• support the existing infrastructure:

Reach 1,000 - 1,500km ( < 10 dB OSNR at 0.5 nm RBW) avoid changes of line equipment, especially current EDFA spacing, and in-line DCF (ideally CD & PMD tolerance comparable to current 10G) > 10 cascaded ROADMs @ 50GHz spacing (LH) > 20 cascaded ROADMs @ 100GHz spacing (Metro)

• Power & footprint same as current 40G (< 40W)• Cost effective to maintain the TCO level

Good Average Bad Good Average Bad0.05 0.1 0.5 0.05 0.1 0.5

5 100 500 1.1 0.9 0.8 1.7 1.1 2.2 11.2 2.0 2.8 11.310 100 1,000 1.6 1.3 1.1 2.4 1.6 3.2 15.8 2.8 3.9 16.015 100 1,500 1.9 1.6 1.4 2.9 1.9 3.9 19.4 3.5 4.8 19.620 100 2,000 2.2 1.9 1.6 3.3 2.2 4.5 22.4 4.0 5.6 22.6

Spans km TotalKm

Fiber Contribution TotalTotal SystemDCUs

System Contribution (All ROADM)AMPsROADM

PMD tolerance increasingly important (given the lack of an effective PMDC):


loukas 12

2323

Outline






24

SP SP EconimicsEconimics

loukas 13

25

Background for the Next-Gen IP Network Revolution



Networks Transition from Service Specific Narrowband Narrowband Access Access NetworkNetwork

Broadband Access Broadband Access NetworkNetwork

Radio Radio Access Access NetworkNetwork

SONET/SDHSONET/SDHAccess Access NetworkNetwork

High Speed (Ethernet)High Speed (Ethernet)Access NetworkAccess Network

Voice NetworkVoice Network(Circuit)(Circuit)

TDM NetworkTDM Network(Circuit)(Circuit)

FR/ATM NetworkFR/ATM Network(Packet)(Packet)

Public IP NetworkPublic IP Network(Packet)(Packet)

Private IP/MPLS NetworkPrivate IP/MPLS Network(Packet)(Packet)

Optical NetworkOptical Network(Circuit)(Circuit)

Challenges:Challenges:•• CapexCapex•• OpexOpex•• Service Velocity Service Velocity


loukas 14

27

Changing Application & Traffic2004 2008

24,500 TB/month 654,000 TB/month

93% CAGR

172,000 TB/month 1,190,000 TB/month

47% CAGR

Busine

ssCo

nsum

er

Rise of Video / IPTV

Proliferation of Business Broadband

Consumer Broadband(TB / month)Consumer VoIP(TB / month)Consumer IPTV / VoDConsumer FTTH(TB / month)

Business DSLIP VPNPrivate Line (IP Portion)

EthernetATM / FR (IP Portion)

Source: Cisco Estimates, Ovum, Bernstein, Public Company Data



Aggregation NetworkAggregation Network

Carrier Ethernet Carrier Ethernet AggregationAggregation

BNGBNG

Business Business PEPE

AccessAccess EdgeEdge

Aggregation Aggregation NodeNode

DSL DSL

Ethernet Ethernet

Core Core

VoDVoD

Content NetworkContent Network

TVTV SIPSIP

Multiservice CoreMultiservice Core

Core NetworkCore NetworkIP / MPLS IP / MPLS

Distribution Distribution NodeNode

STBSTB

CorporateCorporate

STBSTB

STBSTB

ResidentialResidential

CorporateCorporate

CorporateCorporateBusinessBusiness

BusinessBusiness

BusinessBusiness



2G/3G Node2G/3G Node

PONPON

DynamicDynamic or Static Static

DWDM DWDM ROADMROADM

IP/MPLS-over-Optical Next-Generation Networks Wireline & Wireless, Residential & Business, convergence


loukas 15


HR4

ID

F

R1 R2 R3

G

J

CA KE

B

HR4

ID

F

R1 R2 R3

G

J

CA KE

B

VCAT OnlyVCAT Only Packet Based Packet Based VCAT Only CET

91 Ring 1 3084 Ring 2 3049 Ring 3 2621 Ring 4 8245 Total VC4 94

(6) 75% full STM64 Rings (4) 40% full

Required VC4:

Statistical Multiplexing Gain – a simple example

• One 4 site GE Customer• Six PtP FrGE Customers• One SAN Customer: 2GE & 2 FC


30

Router Evolution

.

Late 1980s Early 1992s Mid 1995s 1998Late 1980s Early 1992s Mid 1995s 1998--99 99 20002000--0202 20032003--futurefuture

� scaling the IP POP has been a challenge; interconnect technologies innovation

� Value: Convergence of core, peering, and edge functions = CapEX/OpEx savings

SharedFDDI Ring FDDISwitchX X

POS

EthernetATM

POS POSPOS

GSRGSR

Cisco

75XX

Cisco 75XX

Cisco

75XX

Cisco

75XX

Cisco

75XXDPT

1.2 Tbps to 92 Tbps

Core Routing Systems


loukas 16

© 2007 Cisco Systems, Inc. All rights reserved.© 2007 Cisco Systems, Inc. All rights reserved. Cisco ConfidentialCisco Confidential 3131

Router evolution and the role of optics� Early 90’s: routers over FDDI� Mid 90’s: routers over ATM networks� Later 90’s: routers over SONET � external optical i/fs� Early 00’s: routers over DWDM � direct mapping over optical network – no extra grooming

� Mid 00’s: multi chassis routers � internal optical i/fs between chassis

� Late 00’s: IPoDWDM � DWDM directly in routers� Early 10’s: advanced modulation in routers, interaction with optical switched networks

� Beyond: bigger role for optical processing inside routers?


32

IP/MPLS

ATM / Ethernet

SDH/OTN

� High OPEX unjustified� CAPEX and power higher – spread over multiple technologies� Sensitive to accurate forecast per service type

DWDM

L1

L2

L3

L0

How good is Today’s Architecture for IP evolution?


loukas 17

3333© 2005 Cisco Systems, Inc. All rights reserved.© 2005 Cisco Systems, Inc. All rights reserved.Cisco Confidential, for further details contact Cisco Confidential, for further details contact [email protected]@cisco.com

Benefits of IPoDWDM solution

• Lower CapExElimination of OEOs

• Lower OpExSpace, power, management

• Enhanced resiliencyFewer active components

• Investment protection40G and beyond, interoperability over existing 10G systems

BeforeBeforeBeforeBefore

RouterRouterRouterRouter ROADMROADMROADMROADMTransponderTransponderCrossCross--connect connect TransponderTransponderCrossCross--connect connect

WDM Transponders WDM Transponders Integrated into RouterIntegrated into RouterWDM Transponders WDM Transponders

Integrated into RouterIntegrated into Router

RouterRouterRouterRouter ROADMROADMROADMROADM

DWDM

I/F

DWDM

I/F

34

IP+DWDM Value Proposition

LayeredLayered

Core

Edge

Peering

ROADMWDM

Access

Capex/Opex reduction, Increased Service Flexibility

IP NGN over WDMIP NGN over WDM

0%5%10%15%20%25%30%35%40%45%50%55%60%

0.01 0.1 1 10Relative Network Load

Capex Savings v IP-over-Optical

01020304050

-10% -5% 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50% 55% 60% 65%CapEx Savings v. IP-over-Optical

Freq

uenc

y

IP-over-SONETIP-over-DWDM

IPoWDM Savings (ECOC 2006 W5 – L. Paraschis)


loukas 18

35

Trans-ponder

SR port on router

WDM port on router

Optical impairmentsCorr

ecte

d bi

ts

FEC limit

Working path

Switchover lost data

Protectedpath

BER

LOF

Optical impairmentsCorr

ecte

d bi

ts

FEC limit

Protectiontrigger

Working path Protect path

BER

Near-hitless switch

WDM WDM

FEC

FEC

Today’s protection Proactive protection

IP-over-DWDM Advanced Protection feature Reference: OFC2008 - O. Gerstel et. al. NWD4


36

Experimental Results for MPLS FRRReference: OFC2008 - O. Gerstel et. al. NWD4

TestCase

Proactive FRR Fault

Packet loss (in ms)

Min Max Avg. StdDev.

1 Y Optical-switch 3.10 4.76 4.10 0.7802 Y Fiber-pull 3.17 3.31 3.18 0.0843 Y Noise-injection 0.009 0.134 0.076 0.0444 N Optical-switch 4.51 7.86 5.50 1.0455 N Fiber-pull 3.12 3.41 3.25 0.0936 N Noise-injection 2233 3095 2705 324


loukas 19

37

Open WDM evolution – Alien-Wavelength transmission(Reference: D. Ventorini et. al. Embratel Brazil OFC 2008 NME3 )

BAD FE

IG H

LJ

K

EMS� Definition: Tx/Rx vendor != DWDM net vendor

� Motivation: Avoid costly OEO conversionsFlexible scalingIndependent Tx/Rx innovation

� Management of DWDM network includes Tx/Rx

� Challenge: Missing demarcation between the layers � need good impairment monitoring (OSNR, CD, PMD, …)


38

Open WDM evolution to 40G Alien-Wavelength transmission(Reference: T. Kozar et. al. 2009 Netia Poland )

40G core upgrade http://finance.yahoo.com/news/Netia-Deploys-40-Gbps-Core-to-iw-

14878790.html

40G DPSK IPoDWDM (cisco) 500-700 km links

Over existing 10 Gb/s WDMsystem (Siemens)


loukas 20

39

Next-Gen WDM Transport Goals � More Flexibility: Cost-effective scalable (plug-and-play) transmission and

non-blocking switching

� Manageability: Seamless operations and trouble shooting (w/out requiring a Ph.D. degree)Advanced monitoring of optical impairments for each channel modulation format

� DWDM aware control plane, and intelligent managementRouting based also on WDM (impairment) aware path-computation

� Long-term: Adaptive: collaboration among optical and routing for cost-optimized bandwidth use Future potential for flexible spectrum extraction – not just a fixed grid


Flexible WDM Transport

NJ

Philadelphia

NY

DC

Chicago

Detroit

A

BCD

AB

CD

� Flexible Add/Drop � Directionless (fully Meshed)

� Colourless

� Seamless Provisioning � Routing (regeneration)� Alien Wavelength� Monitoring � Intelligent Control

Plane, and Management

R

Delaware

Baltimore

Indianapolis

loukas© Loukas Paraschis, cisco, 2009. All rights reserved. © Loukas Paraschis, cisco, 2009. All rights reserved.

loukas 21


So far the optical layer was assumed to be static…� Can the optical layer become dynamic?� What value does this bring to the IP layer?� What constraints does the solution have?



Why is the optical layer still not dynamic? 1. Optical switches did not provide switched add/drop until

recently, or were too complicated and expensive2. Network management systems and operations practices

not geared towards dynamic optical networking• Trouble shooting when paths are unknown• Manage resources instead of micro-managing the network

3. Hard to calculate optical feasibility in control plane• Significant optical data exchange• Standard bodies reluctant to burden signaling protocols with

this data• Method for assessing feasibility considered a proprietary “secret

sauce”


loukas 22


Why is the optical layer still not dynamic? 4. DWDM networks have not been built for fast reaction

• Due to optical layer power mgmt • Worse for 40G/100G due to need to adjust Tx/Rx

5. Regenerators in dynamic networks add additional complexity/cost • Need to be pre-deployed to support future traffic • Hard design problem if traffic is unknown

[Raza et al, “Predeployment of resources …”, JLT 2004]6. Lack of clear “killer app”

• Biz case for Bandwidth on Demand (BoD) in DWDM layer hard: must reuse the same bandwidth for enough different users

[Gerstel, “Optical Layer Signaling…” Comm. Mag,, 2000]



Why is the optical layer still not dynamic? (7) Concerns about optical protection

� L0 Protection does not cover all failure modes� Dedicated L0 protection wastes bandwidth� Shared L0 protection relies on unproven and

slow control plane� L0 protection does not deal well w multiple

failures

� Having both L0+L3 protections may create race conditions & oscillations


loukas 23


Signs of Changes – towards Dynamic Optical Networking1. Add/drop optical switches � being productized2. Operations � still not supporting dynamic networking3. Control plane � move towards “impairment aware

GMPLS”4. DWDM networks � will remain slow reacting5. Regenerator pre-deployment � best fix is to remove

them by extending the reach6. Some promising applications are being proposed

Biz case for Bandwidth on Demand � still non-existent7. Optical protection � limited use for now



More Control Plane ExamplesFeature ValueSharing knowledge about degrading links More robust protectionSharing of risk group information from the optical layer

More robust protection

Alarm correlation between the two layers Reduce operational costSharing metrics related to the cost of a lightpaths

Reduce capital cost

Reconfiguration of the optical layer to alleviate congestion on router links

Reduce capital cost

Combined restoration between layers Reduce capital cost

Shar

e Sh

are

info

info

Shar

e Sh

are

actio

nac

tion


loukas 24


Router Driven LightpathsConcept

� Router measures utilization for each link� If utilization is high – request another lightpath from L0� If utilization drops – release unused lightpath



Router Driven LightpathsConstraints

� Must stick to original IP topology� Otherwise – may destabilize IP routing� And create congestion elsewhere


loukas 25


5 nodes, 14 links5 nodes, 14 links28 interfaces28 interfaces5 nodes, 7 links, 5 flex 5 nodes, 7 links, 5 flex i/fsi/fs19 interfaces19 interfaces

Router Driven Lightpaths How does this reduce the cost?Today� Need to over-provision to ensure capacity exists when IP needs it. Need to do it everywhere since location of surge is unknown

Automated IPoDWDM� Extra interface per node can be deployed when a link becomes congested. No need to over-provision per link

Spare i/f Spare i/f --currently unusedcurrently unused

Spare i/f reinforcing Spare i/f reinforcing congested linkcongested link

UnderUnder--utilized utilized linkslinks© Loukas Paraschis, cisco, 2009. All rights reserved. © Loukas Paraschis, cisco, 2009. All rights reserved.


Router Driven LightpathsDiscussion

• Real life traffic surge evidence:• [“Handling IP Traffic Surges via Optical Layer Reconfiguration” Carnegie Mellon/ATT Labs, OFC 2002]• “When we analyze the surge magnitude distribution, we find that the addition of one extra, temporary link between affected router pairs will support up to 97% of all surges...”

• Concerns:• Is it worth pre-deploying 100G Tx/Rx w/o using them?• Is WL granularity too coarse?• How much churn in the optical layer?


loukas 26


Adaptive WDM transmission vision

� Does the network need 100G for every pair of core routers?Not necessarily… Is there a way to cost-optimize component, regen, scaling ?… and still maintain Optical Network simplicity and flexibility…

Network @ 10GNetwork @ 10G Increased use of link Increased use of link bundlingbundling Adaptive NetworkAdaptive Network© Loukas Paraschis, cisco, 2009. All rights reserved. © Loukas Paraschis, cisco, 2009. All rights reserved.


Shifting away from a Rigid Optical Layer� Towards an adaptive network:

1. Adaptive bit rate per channel2. Flexible usage of the Spectrum 3. Sliceable Tx/Rx

� Can this be done?� Is it worth it?


loukas 27

53

Is adaptive Bit-Rate & Spectrum Transmission Possible ?Reference: Paraschis, Gerstel, 2009 LEOS TuD-2.2 invited presentation

50GHz Grid

Flex Spectrum

50GHz 200GHz Future 25GHz

10G2000 Km

100G 500 Km

100G2000 KmNo

n-ada

ptive

100G2000 Km using more spectrum

30G 2000 Km

Adap

tive

100GHz Grid


54

Yoo et. al. “360 Gb/s Optical Arbitrary Waveform Generation” LEOS2008

Research in adaptive Bit-Rate & Spectrum Transmission Willner et. al. “PSK/ASK Variable Bit Rate” ECOC 2006


loukas 28

55

Summary� Network traffic, predominantly IP, is growing > 40-50% CAGR, approaching by 2012 the Zettabyte per year

� Need for cost-effective in scale and operations

� WDM has offered the most scalable transmission and has evolved to an reconfigurable Multi-service Open Transport layer

� WDM transport is scaling 40 Gb/s and will need cost-effective 100G over existing infrastructure asap

� Convergence of IP & WDM offers significant Network benefits, and future Router & WDM integration potential for more sophisticated collaboration towards an adaptive cost-optimized WDM transport

where expensive resources like 40G and 100G could be conserved using advanced monitoring, optical switching, and intelligent network control plane.


56

For questions/comments, please contact:For questions/comments, please contact:Loukas Paraschis Loukas Paraschis [email protected]

Thank you

loukas 29

57

Other Vendors also Believe this is the Future


HardHard

How How

??EasyEasy

How different optical switching technologies relate to each otherTechnology� OPS (Packet)� OBS (Burst)� OFS (Flow)� Agile IPoDWDM� Today’s DWDM

Switching speed� 1e-7 seconds (*)� 1e-5 seconds (**)� 1e-3 seconds� Seconds-minutes (***)� Days-Months (***)

(*) (*) Assuming 1KB packetsAssuming 1KB packets(**) (**) Assuming o(100) packets per burstAssuming o(100) packets per burst(***) (***) Assuming steady state switching (not protection)Assuming steady state switching (not protection)

loukas 30


Current Limitations of Optical Packet Switching

� The important challenge in router scalability:� Reduce cost of transferring a bit through a node� Reduce power consumption

� OPS technical challenges:– Buffering – or else blocking– Cost per optical gate

� Alternative practical evolution in the IP & Optical synergies:� IP-over-DWDM is already reducing cost, power, improve

reliability, maintain scalability, and…� can allow for increased network intelligence…


Vincent Chan Claude E. Shannon Communication and Network Group, RLEVincent Chan Claude E. Shannon Communication and Network Group, RLE

1.1. Abundance of Abundance of bandwidthbandwidth2.2. No costNo cost--effective optical effective optical random access memoryrandom access memory3.3. SiCMOSSiCMOS ~ $10~ $10--88/gate /gate –– optical ~ optical ~ $10$1044--101055/gate, 12 orders /gate, 12 orders 4.4. Min Min switching energyswitching energy of optical logic gate ~ hof optical logic gate ~ hνν, >> , >> kTkT5.5. Electronic time slot interchangers cheaper (~ 6Electronic time slot interchangers cheaper (~ 6--7 orders) than dual in the optical 7 orders) than dual in the optical

WDMWDM domain: domain: wavelength converterswavelength converters..6.6. Optical Optical wavelength switcheswavelength switches are more efficient switching in bulk are more efficient switching in bulk 7.7. Optics inherently a better Optics inherently a better broadcast broadcast medium medium 8.8. LasersLasers are more are more expensiveexpensive than electronic signal sources.than electronic signal sources.

Electronic network architectures are designed for efficient use of electronics Electronic network architectures are designed for efficient use of electronics ––fine grain fine grain switchingswitching

Optical network should use architecture created to exploit optical properties of light Optical network should use architecture created to exploit optical properties of light ––coarse granularity switchingcoarse granularity switching

Differences between optical and electronic technologiesDifferences between optical and electronic technologies

photonics enabling the zettabyte network evolution1122 fault localization in an optical network...

Documents