Olivier Jerphagnon, Calient Networks E-mail:...

“Advances in Photonic Switching & GMPLS”Cisco Optical Technical Workshop19th APAN Meeting, Bangkok, Thailand

Olivier Jerphagnon, Calient NetworksE-mail: [email protected]

2Cisco Systems and Calient Networks Confidential. Covered under Joint Marketing Agreement

Outline

Application Examples

Photonic Switching

Migration to IP/Optical

Q&A

GMPLS Networking


• Founded in March 1999. Two key technologies:3D-MEMS based Photonic Switching

Standard based GMPLS Control PlaneSan Jose, CASoftware & Systems

Santa Barbara, CAOptical Design & Manufacturing

Ithaca, NYMEMS Design & Fabrication

Introduction to Calient

XX

PXC 128/256Carrier-class GMPLS Switching System

PX32 - PX288Optical Fiber Switch

GMPLS ControllerAdding intelligence to non-

GMPLS native devices

DiamondWave® Product FamilyDiamondWave® Product Family

AFM 72-640Automated FiberManagement

XX

PXC 128/256Carrier-class GMPLS Switching System

PX32 - PX288Optical Fiber Switch

GMPLS ControllerAdding intelligence to non-

GMPLS native devices

DiamondWave® Product FamilyDiamondWave® Product Family

AFM 72-640Automated FiberManagement


Future of NREN: Migration from pure IP to hybrid IP/Optical

444© 2004, Cisco Systems, Inc. All rights reserved.

Further reading: Lightwave Magazine (Nov. 2004 issue) on “Photonic Switches in worldwide NRENs”.


Drivers to Migrate from IP to IP/Optical

• ApplicationsDynamic and higher capacity bandwidth services for e-science

• PerformanceLow Latency to allow distributed computingHigh Throughput to maximize bandwidth

• Capacity & ScalabilityScaling from 1 Gbit/s to 1 Tbit/s connections

• It’s also about economics (“paradigm shift”)

Cost of optical transport decreasing faster than cost of routingLeverage access to dark fiber vs. leased circuit

NRN/RE Migration

X.25 ATM POS Optical

1980s 1990s 2000s

Learning, Collaboration, Connectivity

Advanced Services

2Mbps 34Mbps 155Mbps 622Mbps 2.5Gbps 10Gbps nx10Gbps nx40Gbps

Market Transition

?

GRID Applications


Example of Drivers

• Need for higher capacity services for particle physics (ex: CERN LHC)

• Continuing the Trend: ~1000 times BW Growth per decade

• Capacity increase drives the need to learn how to use Multi-Gbpsconnections dynamically, and new technologies that are cheaper and more scalable

• High-end applications (visualization, VLBI, etc) demand specific low latency transport and guaranteed bandwidth

• Cost of photonic switches are 1/10 to 1/100 cheaper than SDH and IP router ports at 10G

Year Production Experimental Remarks 2001 0.155 0.622-2.5 SONET/SDH

2002 0.622 2.5 SONET/SDH DWDM; GigE Integ.

2003 2.5 10 DWDM; 1 + 10 GigE Integration

2005 10 2-4 X 10 λ Switch; λ Provisioning

2007 2-4 X 10 ~10 X 10; 40 Gbps

1st Gen. λ Grids

2009 ~10 X 10 or 1-2 X 40

~5 X 40 or ~20-50 X 10

40 Gbps λ Switching

2011 ~5 X 40 or ~20 X 10

~25 X 40 or ~100 X 10

2nd Gen λ Grids Terabit Networks

2013 ~Terabit ~MultiTbps ~Fill One Fiber

Source: Larry Smarr (UCSD), “Major Links: Bandwidth Roadmap in Gbps”


Hybrid IP/Optical Node Architecture

Photonic switch

DWDM / ROADM

LeasedCircuits

STM-16/6410GE WAN

Lambda

Transport Options

GMPLS Multi-layer Control Plane

Layer-2 Services

Production and Experimental Layer 1/2/3 services

CWDM

Layer-3 Services

Layer-1 Services

10 GE LANGE

10/100 FEGE

Dark Fiber

• Photonic Switches & GMPLS are key elements to address new goals,and implement a multi-tiered and scalable IP/Optical network


Key Applications

• Core NetworksIP core optimization: providing more bandwidth and service capabilities at lower cost

End-to-end services (bandwidth-on-demand) across multiple types of equipment and network domains

Distributed supercomputing (GRID)

• Optical Exchange PointsBandwidth and Service manager

Internet Exchange Providers


Photonic Switching



ASIC Voltage drivers

MEMS mirror array

Collimator array

3D-MEMS Photonic Switching

• 3D MEMS designBulk MEMS

Single crystal hinges

Non-contacting motion

Electrostatic actuation (low power)

Hermetically sealed

Comb-drive (high angles)

Array integration and short path-length allow low cost

Truly transparent (2.5G, 10G, 40G and beyond)

Scalable to large switching capacity (> 10 Tb/s)


Switch Performance

• Low insertion loss (1.5 dB typ.)

• Wide input power range

• Fast switching time (20 ms typ.)

• Truly transparent:Wavelength transparent (1260-1625nm)

Bit-rate, format and protocol independent

• Behaves like Single-Mode-Fiber with extra loss:

Low PDL (<0.2 dB)

Low Crosstalk (<-60 dB)

No non-linearity, no dispersion

0

2400

4800

7200

9600

12000

0 0.5 1 1.5 2 2.5 3 3.5 4

Insertion Loss (dB)

Num

ber o

f Con

nect

ions

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%100%

Cum

ulat

ive

Perc

enta

ge

320x320 connections below 3.5 dB

Transparent from 1260 to 1625nmTypical Wavelength Dependent Loss

Note: the higher loss peak around 1400 nm is due to OH water absortion and can vary.

00.5

11.5

22.5

33.5

44.5

5

1250 1300 1350 1400 1450 1500 1550 1600 1650

Wavelength (nm)

Inse

rtio

n Lo

ss (d

B)

PX

O E S C L

Transparent from 1260-1625 nm


Switch Reliability

• Mature core switching technology

Field proven, shipping for revenue for > 2 years

NEBS compliant, including Zone-4 Earthquake

High switch availability

• Solid MEMS technology

Billions of MEMS mirror cycles without a single failure

Accelerated aging tests shows low FIT < 35

ISO 2001 MEMS foundry

• Thorough environmental tests performed including temperature/humidity, vibration, shock, etc

• Redundant Power-Supply A&B feeds


Vibration & shock isolation

• Thorough shock testing to comply with C.O. deployment. Tests include impact testing and office vibration

• Highly resistant MEMS survive acceleration up to 500 G’s (only 30 G’s are required to survive Earthquakes)


Price/Port

Features & FunctionalityLow

Remote Test Monitoring

Automated FiberManagement

Optical ExchangePoints

GMPLS-basedIP/Optical Networks

High

Low

High

Lab Automation

Photonic Switching Applications


GMPLS Networking



GMPLS

•• MPLSMPLS (Multi-Protocol Label Switching) is the set of extensions to OSPF, IS-IS, and RSVP, CR-LDP to support the the routing of paths (aka traffic engineering)

Intrinsic is the notion of separating the forwarding (data) plane from the control (signaling) plane

Fundamental difference with IP and ATMBut uses same protocols developed by IETF

Data is transported across the network inside a path

•• GMPLSGMPLS is the generalization of MPLS and the realization of the MPλS concept, created by extended MPLS to support nonnon--packet packet paths (λ’s, time-slots, fibers)


The GMPLS Framework

• Segregation of data and control plane • Think of it as the “SS7” of data and transport networks• Standard control plane across network elements

DiscoveryConnection signalingInventory informationManagement information

SONET

Router

ATM DWDM DWDM

OEO

SONET

Router

ATM

OEO

Control Plane

Data Plane

GMPLS Signaling


GMPLS Network Hierarchy

• GMPLS Network and Label HierachyPackets (statistical multiplexing)

SONET

Lambda

Fibers

R0 S2 O3R1 S8 R9 R10O7

R

S

O

P4 P5 P6

P

Fiber LSP4

λ LSP3

Time slot (TDM) LSP2

Packet LSP1

PSC

TDM

λSC

FSC

LSP4LSP3LSP2LSP1

Photonic Switch

OpticalOEO switch

SONET switch/mux

Router actingAs an IP LSR

Fiber

OC-192

OC-12c

500 Mbps froma GBE

Legend:

R0 S2 O3R1 S8 R9 R10O7

R

S

O

P4 P5 P6

P

Fiber LSP4

λ LSP3

Time slot (TDM) LSP2

Packet LSP1

PSC

TDM

λSC

FSC

LSP4LSP3LSP2LSP1

Photonic Switch

OpticalOEO switch

SONET switch/mux

Router actingAs an IP LSR

Fiber

OC-192

OC-12c

500 Mbps froma GBE

Legend:


GMPLS Protocols

• GMPLS protocolsLink Management Protocol (LMP)

Extension to WDM: LMP-WDM

Routing (OSPF-TE)

Signaling (RSVP-TE)

• See it as set of tools to implement various network use cases (dynamic bandwidth provisioning, optical restoration, etc) and deployment interfaces (overlay or peer)


R1

R2

R3

S1

O1

R0

P2

P4

P3P1 O2P5

UNI

UNI UNI

UNI

UNI

UNI

UNI

STSI[NNI]

STSI[NNI]

STSI[NNI]

STSI[NNI]

Photonic Transport Layer(Transparent)

Transport Service Layer Transport Service Layer

PTN

OTN

R1

R2

R3

S1

O1

R0

P2

P4

P3P1 O2P5

UNI

UNI UNI

UNI

UNI

UNI

UNI

STSI[NNI]

STSI[NNI]

STSI[NNI]

ILSI

Photonic Transport LayerTransport Service Layer Transport Service Layer

PTN

OTN

R1

R2

R3

S1

O1

R0

P2

P4

P3P1 O2P5

UNI

UNI UNI

UNI

UNI

UNI

UNI

ILSI[E-NNI]

ILSI[E-NNI]

ILSI[E-NNI]

[E-NNI]


PTN

OTN

I-NNIR1

R2

R3

S1

O1

R0

P2

P4

P3P1 O2P5

UNI

UNI UNI

UNI

UNI

UNI

UNI

STSI[NNI]

STSI[NNI]

STSI[NNI]

STSI[NNI]



PTN

OTN

R1

R2

R3

S1

O1

R0

P2

P4

P3P1 O2P5

UNI

UNI UNI

UNI

UNI

UNI

UNI

STSI[NNI]

STSI[NNI]

STSI[NNI]

STSI[NNI]



PTN

OTN

R1

R2

R3

S1

O1

R0

P2

P4

P3P1 O2P5

UNI

UNI UNI

UNI

UNI

UNI

UNI

STSI[NNI]

STSI[NNI]

STSI[NNI]

ILSI


PTN

OTN

R1

R2

R3

S1

O1

R0

P2

P4

P3P1 O2P5

UNI

UNI UNI

UNI

UNI

UNI

UNI

ILSI[E-NNI]

ILSI[E-NNI]

ILSI[E-NNI]

[E-NNI]


PTN

OTN

I-NNI

Network Deployment Interfaces

• Examples of network interface implementation :I-NNI: OSPF-TE single area, RSVP-TE, LMP, LMP-WDME-NNI: OSPF-TE multi-area, BGP-MP, RSVP-TE, LMP, LMP-WDMUNI: RSVP-TE and LMP


Examples of Applications



Lower cost network infrastructure: S-SINET

Source: Pr. S. ASANO, NORDUnet 2003

14

TokyoPXC

NagoyaPXC

IP Backbone(Quality Control, Policy Control)

IP Backbone(Quality Control, Policy Control)

OsakaPXC

Telescope for VLBITelescope for VLBI

Super ComputersSuper Computers

UsersUsers

•Realization of DataGRID•Data Sharing with low Latency

Equipmentson Liner Accelerator

Equipmentson Liner Accelerator

Fusion Research EquipmentFusion Research Equipment Network StorageNetwork Storage

Example of GMPLS Control applied to DataGRID on SuperSINET

6

SuperSINET Backbone Logical TopologySuperSINET Backbone Logical Topology

P

P P

P TokyoExchange

OsakaExchange

NagoyaExchange

HokkaidoUniv.

TohokuUniv. KEK

TsukubaUniv.

Institute ofMaterialReserch (Tokyo Univ.)

WasedaUniv.

TokyoInstitute ofTechnology

NII Chiba

NII Hitotsubashi

NAOTokyoUniv.ISAS

NIG

OkazakiNationalResearchInstitutes

NagoyaUniv.

NIFS

DoshishaUniv.

KyotoUniv.

(Yoshida)

KyotoUniv.(Uji)

OsakaUniv.

KyushuUniv.

Institute ofMedicalScience(Tokyo Univ.)

PE

PE

PE

PE PE PE

PE

PE PE

PE PE PE

PE PE PE

PE

PE

PE PE PE

PE

PE

BackboneOC192

Customer AccessOC48,GbE

Cisco12400

Customer Router

• “Economical Solution for High-performance: Dark Fiber (6,000Km+) with 300G+ bandwidth including Transport Equipments (DWDM, CWDM, 3R Repeaters, Amps) and PXCs for 3 years contract will realize 90% cost reduction compared to equal bandwidth of SDHs


End-to-end Bandwidth-on-Demand services: JGN-2

L2SW

Fukuoka

Osaka

STM-64

GbELX

Kanazawa

Otemachi

P XC

GbE

MU

X

WD

M10G

NECMG4040

GbELX

GbE MUX

WD

M

GbE

MU

X

GbE

MU

X

GbE

MU

X

P XC

L2SW

L2SW

STM-64

STM -64

STM-64

STM-64

STM-64

STM-64

GbE S XGbE LXSTM -6410GE(W): 10GbE WAN-P HY(LW )10GE(L): 10GbE LAN-PHY(LR)

GbELX GbE

LX

P XC

MC MC MC

E-NNI

WDM-LMP WDM-LMPL2SW

Fukuoka

Osaka

STM-64

GbELX

Kanazawa

Otemachi

P XC

GbE

MU

X

WD

M10G

NECMG4040

GbELX

GbE MUX

WD

M

GbE

MU

X

GbE

MU

X

GbE

MU

X

P XC

L2SW

L2SW

STM-64

STM -64

STM-64

STM-64

STM-64

STM-64

GbE S XGbE LXSTM -6410GE(W): 10GbE WAN-P HY(LW )10GE(L): 10GbE LAN-PHY(LR)

GbELX GbE

LX

P XC

MC MC MC

E-NNI

WDM-LMP WDM-LMP

• GMPLS signaling provides bandwidth-on-demand between edge L2/L3 devices across multi-domains (KDDI and NTT)

• GMPLS allows tiered protection/restoration services and production-grade monitoring (BER monitoring, etc) with LMP-WDM

Source: Tomo Okani, KDDI

More on JGN-2 at Lambda Networking BoF, Wedn.


Distributed Supercomputing: Caltech

• Photonic Switches and GMPLS are ideal to support clusters of supercomputer clusters.

• Connection between supercomputers can be established locally (campus) or remotely (nationwide backbone) via exchange points

To WAN

GBE SXGBE LX10 GbE LAN or WAN PHYOC-48 or OC192

DWDM

MSPP


Optical Exchange Points: UltraLight

• Ex. of one Optical Exchange part of UltraLight project

• L1/L2/L3 services managed by OOO switches integrated with GRID middleware (AAA, MonALISA, etc)


Internet Exchange Providers (IXP)

• Main drive is to improve resiliency and support higher capacity• Migration from ring to double-tree IX architecture with dual Photonic

Switch configuration• Photonic Switches address spanning tree issue and higher scalability• Migration can from old to new architecture can be done quickly once

switches are in place; reverting to previous configuration is possible

10GbEISP AISP B

ISP CISP D

ISP EISP F

GbE

ISP A ISP B ISP C ISP D ISP E ISP F

ISP HISP G

Ethernet Switch

10GbE Switch


Q&A



Control Plane Standard Activities

• Question:What are the roles of ITU, OIF, IETF and GGF? Isn’t there some overlap (ex: E-NNI)?

• Answer:IETF: Develops GMPLS protocols + layer-3 perspectiveITU: Provides architecture requirements (ASON) and layer-1 operation perspectiveOIF: Develops implementation and addresses interoperability considerations for carriersGGF: develops Grid technology via “best practices”, technical specs, user experiences, and implementation guidelines.

They all play their role for the deployment of next-generation networks and provide the infrastructure for new GRID services

They have liaisons to coordinate efforts. Early commercial deployments (ex: KDDI) to set de-facto standards


Optical and Photonic Cross-Connects ?

• Question:What is the difference between an Optical-Cross-Connect (OXC) and a Photonic Cross-Connect (PXC)? Are they competitive?

• Answer:Optical XC is somewhat a misnomer as core switch fabric is electrical (O-E-O). Hence term Photonic XC to differentiate all-optical switch (O-O-O ).

They are complementary as they have different strengths:

OXC/MSPP: Grooming capability. Ideal for legacy SONET/SDH and sub-wavelength (OC-12, GbE, etc) traffic aggregation

PXC: Lower cost and high scalability (ex: up to 320 10G ports). Support true IP/Optics with 10GbE LAN-PHY switched over DWDM wavelength at 10th cost of switched SDH managed service. Future proof (40G ready, etc)


Photonic Switches and ROADM

• Question:Both Photonic Switches and ROADM provide some level of optical swicthing. Are they competitive?

• Answer:No. In the contrary they are complementaryROADMs: Integrated DWDM and cost optimized for rings, Limited switching capability (< 32x32)Photonic Switches: Larger switching capability (>32x32); Complementary functionalities including ring interconnect, mesh optical networking, optical multicasting, etc

11

11

ROADM

11 11

1111

11

11

11Interconnect ring site

To National backbone

Metro Ring 1

Metro Ring 2

Ex: US Regional Optical Networks

Olivier Jerphagnon, Calient Networks E-mail:...

Documents

Transcript of Olivier Jerphagnon, Calient Networks E-mail:...