© 2003 – rnp experimental optical networking: implementations and challenges internet2 spring...

© 2003 – RNP

Experimental Optical Networking: implementations and challengesInternet2 Spring Members’ Meeting, Arlington, VA, USAApril 2003Michael StantonRede Nacional de Ensino e Pesquisa do Brasil - RNP<[email protected]>

Michael Stanton - Experimental Optical Networking - I2MM, April 2003

2

What do we mean by Optical Networking?

• End to end optical routing?– Not yet visible on our horizon

• High capacity national and international pipes?– Already available, permitting support for arbitrarily

large and complex applications– competition, declining costs

• User control of networking infra-structure?– Alters the user dependence on costly telco services– New paradigm for future of networking


3

Traditional networking model - “service-based”

• Involves setting up IP infrastructure based on leased telco services

• Telcos were state monopolies until recently in most countries, and still are in some:– no competition

• arbitrary (high) pricing• unavailability of some (advanced) services

This can be an opportunity to influence the telco offerings through collaborative innovation!


4

Replacing copper by optical fibre- “facilities-based”

• Technical advantages:– (much) greater transmission capacity– (much) greater distances between equipment– (much) lighter and cheaper

simplifies the infrastructure for networking makes feasible user ownership or provision

• This can be be put to use at many levels:– corporate, access, MAN, WAN


5

Infrastructure -how to get it

• Rent or borrow it (long term IRU contracts)– current glut of fibre in many places

• Condominium fibre• Build it out yourself

– SURA optical cookbookhttp://www.sura.org/opcook


6

Optical transmission- attenuation in (standard) optical fibre

• Depends on the wavelength (colour)

• Two windows of minimum attenuation in 1310 and 1550 nm bands (infra-red)

• 0,2 dB/km (50% in 16km)

• 50 THz width => 50 Tbps theoretical capacity

low attenuationwindows (50 THz)


7

How to use this 50 Tbps capacity?

• Traditionally terminate optical circuits with electronic switches (OEO) - limitations of electronic switching– IP routers up to 10 Gbps– SDH switches up to 40 Gbps

• Alternatives for greater capacity:– Optical switching (see later) – WDM (wavelength division multiplexing)

- like FDM, is analogical - channel usage depends on transmitter signalling rate

• CWDM (coarse WDM) - fewer channels, more widely spaced => cheaper lasers, shorter distances

• DWDM (dense WDM) - many more channels, closely spaced, dearer lasers, broadband amplifiers


8

DWDM - Dense Wave Division Multiplexing

• Note: we still need to terminate each lightpath with electronic equipment

TE

TE

TE

TE

TE

TE

TE

TE

TE

TE

L

L

L

L

L

L

TETerminal

Equipment

WDM transmitter

(laser)

BroadbandAmplifier

Multiplexor-demultiplexor

Optical components:

Optical line amplifiers


9

Some capacity records for DWDM

• 1×40 Gbps up to 65 km (Alcatel’98) 40 Gbps• 32× 5 Gbps up to 9300 km (1998) 160 Gbps• 64× 5 Gbps up to 7200 km (Lucent’97) 320 Gbps• 100×10 Gbps up to 400 km (Lucent’97) 1 Tbps• 16×10 Gbps up to 6000 km (1998) 160 Gbps• 132×20 Gbps up to 120 km (NEC’96) 2,64 Tbps• 70×20 Gbps up to 600 km (NTT’97) 1,4 Tbps• 80×40 Gbps up to 60 km (Siemens’00) 3,2 Tbps• 1022 lambdas in a single fibre (Lucent 99)• 64×40 Gbps up to 4000 km (Lucent’02) 2,56 Tbps


10

Optical Switching

• General problems:– switching speed (depends on opto-electronic or opto-

electromechanical technology)– header processing speed– temporary storage (for buffers)

• Three major granularities for switching:– wavelength (lambda) switching [circuit switching]

• static v dynamic (i.e. management v signalling)– optical packet switching (OPS) [like electronic PS]– optical burst switching (OBS) [less demanding than PS]


11

Wavelength switches in WDM networks

(a) OADM(b) OXC (without conversion)(c) OXC (with conversion)

1 2 3 4 55

4

32

1

1 2 3 4 55

4

32

1

5

4

32

1

2 2

(a) (b) (c)

A A

B

BB

A AA A

B


12

Optical Packet Switching (OPS)

Electronic processing of header bits Temporary data storage is needed (fibre delay lines - FDL)

(From IEEE Communications Magazine)


13

Optical Burst Switching (OBS)

Packets buffered at network edge; burst control packet sent ahead to“prime” the switches; burst of packets sent end-to-end; no intermediate buffering

(From IEEE Communications Magazine)

Does not read/process (control) bits at extremely high speed


14

IP over optical

• transition to simplified model– 2 level model

IP/WDM

• packet/burst switching - core network - GMPLS• lambda switching

– core network approach - GMPLS– customer network approach - OBGP (Canarie)

FIBREWDM

SONET/SDHATM

IP/MPLS


15

State of the art of optical networking

• NSF classification of networking testbedsbeyond Internet 2 (Tom Greene)– Experimental Infrastructure Networks (EIN) - Internet 3– Networking Research Testbeds (NRT) - Internet 4

• Internet 4 optical networks– dynamic lambda-switched, OPS, OBS networks

• Internet 3 optical networks– Based mostly on statically lambda-switched networks– growing number of networks providing production

networking support for advanced applications


16

Some current experimental optical networking projects in Latin America

• Chile:G-REUNA - Advanced Applications Testbed

• Brazil:GIGA Project - Optical Networking and Applications Testbed

Both of these are a mixture of EIN and NRT (Internet 3 and 4)


17

G-REUNA (Chile) experimental network (2002-2003)

• Phase I of G-REUNA:• R&D in optical

networking and advanced applications

• IP/DWDM• govt. and telco support

• Phases II and III: Deployment of National Backbone: August 2003-July 2004

• http://redesopticas.reuna.cl


18

GIGA Project (Brazil)

• Partnership between RNP and CPqD (former telco monopoly’s R&D centre in Campinas, SP) www.rnp.br and www.cpqd.com.br

• Explore user control of optical fibre infrastructure– interconnect 20 academic R&D centres along the corridor

Campinas - Rio de Janeiro (~600 km)– use of IP/DWDM with Ethernet framing

• Support R&D subprojects in optical and IP networking technology and advanced applications and services

• Industry participation (telcos provide the fibres; technology transfer of products and services required)

• Government funding for 3 years - started December 2002


19

Geographic localisation of network

Campinas

São Paulo

São José dosCampos

CachoeiraPaulista

Rio de Janeiro

Petrópolis SalvadorRecife

Curitiba

Phase I

Phase II

Niterói

FapesptelcosUnespUSP - H.C.USP - C.U.

CPqDLNLSUnicamp

LNCC

CPTEC

UFF

CTAINPE

CBPF-LNCCFiocruzIMEIMPA-RNP

PUC-Rio telcosUERJUFRJ

UniversitiesIMEPUC-RioUERJUFFUFRJUnespUnicampUSP

R&D CentresCBPF - physicsCPqD - telecomCPTEC - meteorologyCTA - aerospaceFiocruz - healthIMPA - mathematicsINPE - spaceLNCC - HPCLNLS - physics

(About 600 km extension - not to scale)


20

GIGA Project:Initial design of the network

• DWDM WAN between Campinas and Rio de Janeiro• Optical MANs in Rio,

S. Paulo and Campinas• Switches between WAN

and MANs for IP packets and lambdas (under study)

• campus networks a problem:not ready for GigE - no SM fibre

Petrópolis

CPLSJCSPO

T-DWDMT-DWDM

AD-DWDM AD-DWDMAD-DWDM

Switch

Switch SwitchSwitch

Switch

T-AOL

CPSRJO

Niterói

CPS

SPO

SJCCPL

RJO

CPS - CampinasSPO - São PauloSJC - São José dos CamposCPL - Cachoeira PaulistaRJO - Rio de Janeiro


21

International Optical Connectivity for Latin America

• Latin American NRENs and Internet 2 access:– some now via AmPath or direct (Mexico)– more (soon) via CLARA (end of 2003)

to GÉANT and ABILENE

• Next step should be to enable Internet 3 lambda connections to the global networks– this would permit interoperation with collaborating

optical networks in other countries– support for network intensive science to participate

at the international level


22

Conclusion

• Optical infrastructure makes possible and desirable new ways of building advanced networks

• Several basic optical switching technologies are under development in testbed networks

• For the present, most experimental infrastructure networks are based on lambda-switching, but this may well change, at least in core networks

• Requirement to link Latin American optical networking to global initiatives

© 2003 – rnp experimental optical networking: implementations and challenges internet2 spring...

Documents