100G

Dr. Paul BrooksMay 2012

JDSU 40G, 100G & BeyondDeployment Challenges

© 2011 JDS Uniphase Corporation | JDSU CONFIDE NTIAL AND PROPRIETARY INFORMATION

Aim

� This presentation aims to help you understand the following:-– Where we are today with 100G (40GE,

100GE & OTU3/4)– Optics & the CFP– The challenges towards successful

deployment– The future evolution & revolution


JDSU – an introduction

Annual Revenue

Business Segments

Employees

Locations

Country Representation

Index Membership

$1.3B

Leader in 3 key segments

≈ 4,100

Over 80 Sales and R&D sites globally

164 Countries

S&P 500


Global Leaders in the Markets We Serve

Telecom, Datacom, Submarine, Long Haul, Metro, Access, Biotech, Microelec

Communications & Commercial Optical

Products

Advanced Optical Technologies

Currency, Defense, Authentication, Instrumentation

Communications Test & Measurement

Telecom/Cable Access, Metro, Core & Home Networking


Challenges for 100G and beyond

� Every generation of communications technology as it’s own challenge!


� Core traffic is doubling every 18 months BUT the change in nature of this traffic is far more disruptive.

� Driven by video– In the US, operators report that over 20% of Internet backbone traffic

is Netflix– Whether fixed or mobile, video growth shows little sign of slowing– Real time & mobile, latency is becoming important

6

Our Environment

2010 2011 2012 2013 2014 2015

Gbytes

Capacity

Demand

Source: Infonetics Research, empirical estimates


What’s happening in the world of 40G/100G?

� Service providers trialing solutions from equipment vendors� Equipment manufacturers continue 40G/100G development

– 40GE, 100GE– OTU3 with 40GE client, OTU4 with 100GE client, OTU4 with

multiplexed ODUk– STL256.4 for 40G SDH/SONET

� Chip vendors developing 100G related products� Transponder manufacturers introducing new products

– More CFP products emerging– Smaller transponders for 40GE----QSFP+– Next generation of transponders---CFP2– Line side modules for 40/100G coherent DWDM

7


40G/100G connections

� Two types of traffic:– Ethernet----40GE, 100GE– OTN----40GE/100GE over OTN or Multiplexed OTN signals

� Client interfaces are standardized, line side interfaces are often proprietary

Router � Router via 100GE

Router � DWDM Client IF

OTU4 or 100GE over OTU4

CFP optics 4 x 25G

100G Line side Modulation & Strong FEC

DWDM equipment

RouterRouter Router � Router

CXP/QSFP+

GESONET/SDH

10GEODUk

FC

OTU4 Client IF

4λ CWDM 13xxnm


100G Optics

100G

SR10 MMFCXP

LR4 SMFCFP

100GEData CentreEnterprise

100GE& OTU4

10λ SMFCFP

Line SideOIF MSA

DWDMModulation

100GE& OTU4

• 100G Client Optics is based on CFP• Industry looking to shrink CFP size over time

1550 nm

1310 nm

850 nm

Main currentapproach

CFP22012 Beta

CFP42014

Next StepCFP2 to replace CFP over time- Same photonic interface (LR4)- Electrical bus moving from 10G to 25G

CFP4 to be more of a datacom form factor- Increase port density on routers


100G Ethernet Interfaces

Interface Reach Medium Parallelism Connectors

100GBASE-LR4 10 km SMF 4 λ / dir LC or SC

100GBASE-ER4 40 km SMF 4 λ / dir LC or SC

100GBASE-SR10 100 m125 m

OM3 MMFOM4 MMF

10 fibers / dir MPO

100GBASE-CR10 7 m Twin-axial electrical

10 cables / dir MDI

IEEE 802.3ba

• Main applications:• LR4: Most common interface by far in telecom• ER4: early stages, much smaller market • SR10: Mainly a data center interface (lower cost)• CR10: Very short reach, not used yet• There also exists a non-standardized i/f which would be ‘LR10’


LR: The Main 100G Client Interface

1295.56 1300.05 1304.58 1309.14 nm

100GBASE-LR4- Most common i/f- Telecom - 1310 nm range- IEEE standard

100GBASE 10λ- Used in data centre- 1550 nm range- Santur/Neophotonics

1523 1531 1539 1547 1555 1563 1571 1569 1587 1595 nm

High-speed Client Interfaces are multi - λ

4.5 nm

8 nm

4 X 25G

10 x 10G


40G Optics

40G

SR4 (100m MMF)QSFP+

LR4 (10km SMF)& G.695 CFP

40GEData CentreEnterprise

40GE& OTU3& OC-768/STM-256

300-pin MSA

OC-768/STM-256/OTU3/OTU3e1/3e2

QSFP+2011

FR (2km serial SMF)& G.693 CFP

40GE & OTU3 & OC-768/STM-2562011/2012

Line SideOIF MSA

DWDM2011+

• 40GE optics moving to QSFP+ � higher density, lower cost• Serial CFP (FR) replacing 300-pin MSA for clients

1550 nm

1310 nm

850 nm

1550 nm

LegacyNew Wave

BecomingAvailable

4 λ 1 λ 1 λ

Being Replaced


40G Ethernet Interfaces

Interface Reach Medium Parallelism Connectors

40GBASE-LR4 10 km SMF 4 λ / dir LC or SC

40GBASE-FR 2 km SMF Serial LC or SC

40GBASE-SR4 100 m125 m

OM3 MMFOM4 MMF

10 fibers / dir MPO

40GBASE-CR4 7 m Twin-axial electrical

10 cables / dir MDI

IEEE 802.3ba and 802.3bg (FR)

1271 1291 1311 1331 nm

40GBASE-LR4λ are different from 100GBASE-LR4

20 nm


Current Trends

� First generation products– CFP based– FPGA

� Need to get 100G client to primetime!– Economics (price vs. 10G)

• QSFP+ is getting there• CFP=> no, likely met with CFP4?

– Port density (very important for enterprise)– FPGAs enabled technology – ASICs deliver port density & price point for primetime

� CFP2 – 25/28G I/O

� Inter-op headaches at all levels– From MDIO & I^2C to FEC & OTN muxing

Client Interface


A 4x25G CFP

CFP module– photograph courtesy of Opnext


CFP Module – Block Diagram for Gearbox CFP

For 100G M=10 & N=410 electrical lanes running at 10G4 optical lanes running at 25GREFCLK is electrical lane speed/16 ~644 MHz


CFP Optics – Current Sitrep

� Products available, early production grade– First series parts – some performance caveats

� Gearbox very challenging technology– Other vendors announced gearbox ICs (CMOS)

� Longer term gearbox will be integrated in line card ASIC to make simpler 4x25G electrical i/o (CFP2)– Quad 25/28G CDRs becoming available– CFP is not the end game for 100GE optics

� 40GE - 4λ parts are more ‘available’– Very aggressive price point for LR4 ($1k) within ~2 years– Will migrate to QSFP+ LR4 through 2012

� First demos of CFP2 expected 2012


100G CFP Migration Path

The form factor for 100G optics will shrink driving higher levels of photonic integration. This is driven by faceplate d ensity and price-point expectations.

Port expander10 x 10GE


100G Form Factors & faceplate density

400Gb

800Gb

1600Gb

3200Gb

Line SideDWDM Transport


Upgrading Networks to 40 /100 Gbps Data Rates

Transmission

Transport

Goal:More bandwidth/capacity over installedinfrastructure(with existing fibers, EDFAs, DCMs, and OADMs)

Solution:Higher data rates andimproved spectral efficiency

Transport Impediments:Optical Noise (OSNR)Bandwidth NarrowingDispersion (CD, PMD)Nonlinearities (SPM, XPM)

Enabling Technologies:Advanced FEC CodesNew Modulation FormatsAdaptive Dispersion CompensationCoherent Detection / Electronic DSP

Upgrading Systems to 40 Gbps or even 100 GbpsFaces Serious Transport and Transmission Issues


Intradyne Coherent Receiver with Real-Time DSP

Input

LO

90o

90o

|| - Polarization

| - Polarization

PBS I

I

Q

QHigh-SpeedDigitalSignalPro-

cessor

Intradyne Coherent Receiver with Phase and Polarizatio n Diversity� Frequency-locked local oscillator laser (no phase-locking needed)� Detection of in-phase and quadrature phase signal co mponents

• Input signal is mixed with 0o- and 90o-shifted local oscillator (in optical hybrid)• Phase offset between signal and local oscillator is removed electronically in DSP

� Detection of polarization components parallel with a nd orthogonal to local oscillator polarization state (no polarization matching needed)

• Multiplexed polarization states are separated electronically in DSP


Block Diagram of Signal Processing Steps

Input

LO

90o

90o

|| - Polar.

| - Polar.

PBS I

I

Q

QHigh-SpeedDigitalSignalPro-

cessor

ClockRecovery

ClockRecovery

ChromaticDispersion

Compensation

ChromaticDispersion

Compensation

PMDCompensation

/ Polarization Demultiplexing

PMDCompensation

/ Polarization Demultiplexing

CarrierPhase

Estimation

CarrierPhase

Estimation

SignalDecoder

SignalDecoder

DSP Functional Block Diagram

Powerful electronic dispersion compensation solves PM D problemand removes need for in-line CD compensation.


OIF Line Side for 40G, 100G and Beyond

� OIF driving an MSA for 40G and 100G line side modules� Define a common form factor, mechanics, connector, power

etc – Module vendors can have their own secret sauce in modulation

& soft FEC� 5 in. x 7 in. module at 80W (2.3W/ in^2)� Connector is Hirose 168pin MDIO is basis of control i/f� For details please visit : OIF� First modules being trialled now and showing much promise.� Power dissipation and other performance issues being

addressed with 2nd generation parts.


2008 2009 2010 2011 2012 2013 2014 2015 2016

100G - Longer Term Outlook

Data earlier – transport later

Proof of Concept phase

100GE IEEEstandardized

JDSU deliver 1st

100G tester 100G will overtake40G spending

Transition to QSFP 40GE?

100G technology becomes

‘mainstream’

1st OIF modules – 40G

100G coherent OIF modules

2nd Gen 100G appear


Product Realization Phases

� R&D phase– First generation (CFP) & FPGA - done– 2nd generation (CFP2 & ASIC) – underway

� SVT– Currently underway and increasing in scope– Deeper OTN– Enhanced functionality (syncE)

� Production– Starting now, focus on throughput & price– Often with contract manufacturer

� Field deployment– Focus on trials and service turn up


Current Technology Challenges

� CFP2/4 MSA to define next generation 100G pluggable – Took longer than expected– But flexible (port expander, reverse gearbox etc)

� 25G ICs– Still ‘premium’ parts– Issues with signal integrity– Large FPGA fabric speed and general purpose 400G MAC– Beyond 25G I/O (50G, PAM, photonic on IC?)– Raw bandwidth => stay on chip

� 25G photonics– PICs– EML => DML migration (eye quality?) & inter-op

� Power & power density� Price expectations

– 4 x 10G => sub $1k / client port


What to test (and when)?

� Did I design it right?– Classic R&D phase

� Will it work properly?– SVT

� Can I build it?– production

� Service experience– Service deployment

� Troubleshooting– Field support– Quality of Service


Classic Lab 100G Challenges

Key requirement in early stages:� Transponder Characterization� Line card – FPGA, ASIC & Photonic

integration

Evolution of 100GE services:� System Validation: PCS Layer Validation� 40GE & 100GE QoS� Transport IP applications

OTU4 development� OTU4 Interworking Tests� Framer, FEC & mapper validation� ODUflex & G.HAO


Summary

� 1st Generation 100G established– Premium product– Driven by raw bandwidth need

� 2nd generation under development– Density and price attractive– Will take 100G to primetime

� 40G very cost effective in many applications

� 400G under development

� Transition for 10G I/O to 25/28G I/O

� Many challenges ahead


100G – Quiz time!


A quiz!

� Which of these has the highest power density?


Answer

� A 100G router is approximately the same as a cooking oven!

Core Router8.6 kW /m^3

Electric oven9.1 kW /m^3

Sauna0.5 kW /m^3


Back to the Sauna - Why does this matter?

� Power & heat will become limiting factors in many high density/high speed applications

� 8 x CFP2 = ~100W just for the pluggable photonics– Could be driving ~400W per card of cooling

� Equipment will need to go into lower power modes and then very quickly get back on line.

� Service providers hitting hard limits on power m^2 in switchroom.

� They have very large electricity bills!


Quiz #2

� Which of these is the smallest?

A germ Skin depth of 25Gbs signal on PCB trace

Wavelength of red light


Quiz #2

� Which of these is the smallest?

A germ ~10um Skin depth of 25Gbs signal on PCB trace~580nm

Wavelength of red light ~650nm


Quiz #3

� What percentage of peak Internet traffic does Netflixconsume?


Quiz #3

� What percentage of peak Internet traffic does Netflixconsume?

Source: Sandvine Inc. Oct 2011Btw: Netflix HD movie feed is 4800 kbs

100G

Documents

ge client

proprietary information

ge optics

g line

ge otn

ge otu41310 nm

g client optics

g cfp4