enabling dynamic, rapid, and effective end-to-end bandwidth from metro edge to waters edge

Enabling Dynamic, Rapid & Cost Effective End to End bandwidth from Metro Edge to Waters Edge

Chris Liou, VP Network StrategyInfinera

What I want you to take away today

2 © 2011 Infinera Corporation.

Bandwidth elasticity is paramount to deliver services cost effectively wherever and whenever needed

Network flexibility must adapt in real-time to changing requirements, while optimizing efficiencies & minimizing cost

Automation is mandatory to ensure a long term sustainable & scalable OpEx model

Transformative Network Influences


Cloud computing & Virtualization• Data Center transformation

• XaaS adoption

Content & Video

Mobile Internet

Ethernet convergence

Content Provider networks

Network simplification: convergence & consolidation

Bandwidth is changing the way networks are built

Subsea

• Scalable capacity• Optimal $/bit• Future-proof for NG-

services• Integrated back-haul

Regional/LH Core

• Maximize fiber capacity• Optimizing optical reach• Any service - Any where• Bandwidth management• Automation/Ease of use

Metro Access/Core

• Minimizing cost/service• Broad service capabilities• Bandwidth management

and service aggregation• Service Flexibility

From Access to Subsea, BW demand requires architectural re-validation!

Long-HaulRegionalMetro Core SubmarineMetroAccess


Emerging Subsea/Terrestrial Networking Requirements

SLTE

SubmarineTransport

New Services

TerrestrialBack-Haul

Service(eg: 10GbE LAN Phy)

Maximum spectral efficiency

Trans oceanic reach

Simple Dispersion Compensation

Software tunable modulation

Consolidated back-haul + SLTE

Integrated switching

Terrestrial optimized capacityand reach

10GbE LAN and WAN PHY (today)

OC-768 and STM-256 (today)

OTU1 and OTU2

100GbE and beyond (future)

SLTE

More Capacity, New Services, Faster, and at Lower Cost

New Services &End-to-end circuit mgmt

SLTE Simplification

Robust Demand for Subsea Bandwidth


Most significant growth areas: Africa, Latin America, Middle East (>80% CAGR)

Cloud Computing Bandwidth Dynamics


Virtualization & Cloud are driving distributed computing• Large-scale pooling & virtualization requires VM mobility, data set migration, etc.

Cloud network infrastructure is migrating towards mega-data centers• Driving higher bandwidth to & between cloud Data Centers

Cloud computing elasticity drives more dynamic bandwidth• Enterprise to Data Center, Data Center to Data Center

Evolving network dynamics & growth raises many concerns• Scale

• Performance (latency, jitter), QoS

• Security

• Affordability

Is the conventional approach scalable?

Essential Cloud Characteristics (NIST)

8

On-demand self-service

Broad network access

Resource pooling• Location independence

Rapid elasticity• Agility = dynamic allocation of IT capabilities

• Responsive = bring resources to bear when needed (and vice versa)

Measured service

Dynamic resource allocation optimizes IT usage to fit fluctuating business cycles.

Bandwidth Agility & Elasticity are Essential

Virtualization creates unpredictable demands on core network• Leverage IT resources wherever & whenever available

• Today: within data center• Tomorrow: across the network

Elasticity creates challenges for underlying network• Cost-effective scaling (capex & opex)• Networking efficiency (idle bandwidth = wasted bandwidth)• Rapid delivery of low-latency bandwidth, when and where needed• Resiliency for protection against failures

Bandwidth agility & elasticity are best addressed through dynamic bandwidth management in transport layer

Requirements for Optimizing Global Connectivity


Bandwidth elasticity• Instant, flexible service connectivity, anywhere to anywhere

• 1G to 100GbE

• Readily available optical capacity for on-demand bandwidth

Network flexibility & efficiency • Any service, anywhere, on-demand• Minimal stranded capital

Network simplification & cost optimization• Vertical (multi-layer) convergence & optimization• Horizontal (end-to-end) optimization

Automation for enabling dynamic bandwidth allocation

Bandwidth Elasticity through Virtualization

11© 2011 Infinera Corporation.

What is Virtualization?

Example: Server Virtualization

Generic PCs Digital logic Software

Cluster computer

Many applications can share the same pool of resources seamlessly,

enabling more efficient resource utilization.

Bandwidth Virtualization:

Shareable Cloud Bandwidth

DWDM WavesDigital logic(Digital Switching) Software

Services see: Pool of reliable, reusable, “colorless” bandwidth

Digital Optical Network

Bandwidth Virtualization:

Decoupling Services and Transmission

200G available 165G available

150G available80G available

Bandwidth Virtualization

40GbE

100GbE

110G available40G available

100G available 65G available

100Gb/10Gb/

Optical Wavelength Agnostic Services Layer

Bandwidth VirtualizationLayer

OpticalCloud SwitchingLayer

•Any service speed over any optical infrastructure •Enables rapid digital service delivery & on-demand reconfigurability

Digital switchingDWDM Waves Software

Dynamic Pool of Bandwidth

Benefits of Bandwidth Virtualization

Without Bandwidth Virtualization, Capacity becomes “stranded”

With Bandwidth Virtualization

• Muxponder inefficiency strands sub-λ bandwidth

• Bandwidth management allows full capacity to be used for services

Used capacity

Constrained capacity

Flexible capacity

Bandwidth Management Enhances Network Efficiency & Responsiveness

Service demand

• Wavelength Blocking strands wavelengths on spans

Colorless BW

Service demand

• Wavelength conversion at every node

Implications of Bandwidth Virtualization

Optical layer can be designed & optimized for best economics• Design each network link for optimal reach + capacity

• Eliminate over-engineering

Bandwidth Virtualization abstraction optimizes utilization of network assets• Pooled wavelength capacity as a shared resource can be utilized by any

service

• Investment protection for deployed infrastructure• 100GbE, 400GbE, 1TbE service readiness

Manage circuits & services, not wavelengths• Per-circuit provisioning, redirection, protection/QoS

• Dynamically allocated bandwidth for services faciliates bandwidth on demand

Network Flexibility & Efficiency through Digital Optical Networks


Nodal bandwidth management can generally be achieved through:• Optical re-direction (1 or more wavelengths)

• Digital switching (converts optical bandwidth into digital domain before switching)

Key Issue: what size circuits am I delivering?

Network Flexibility Approaches


• 2 degree PLC• Multi-degree WSS• CDC ROADM (Colorless,

Directionless, Contentionless)

• Digital ROADM• OXC• Digital/optical hybrid

All-optical approaches Digital Approaches

Conventional DWDM ROADM Architecture


New services are added optically

Wavelength capacity is constrained between 2 endpoints

Trans-mission

Fiber

Dispersion

comp.

(R)OADMN-stage

Filter or WSS

10G

40G

100G

Amp

Transponders or

Muxponders

4-Degree ROADM

Traditional ROADMs create complexity

West East North South

. . . 1 2 39 40

. . . 1 2 39 40

. . . 1 2 39 40

. . . 1 2 39 40

Requires: • Splitter/WSS for every degree• Physical cabling between cards• Physical changes to change

channels• Expensive to create true

flexibility of• Colorless • Directionless &• Contentionless (if available)

The complexity of Optical ROADM’s require optical paths from every point that needs optical flexibility,

requiring, in some cases, over 100+ fiber connections

Traditional ROADMs also lack flexibility


Node Constraints

• Transponders tied to a specific direction and color

• Cannot support 25 GHz

• Prevents dynamic reconfigurability or restoration

O-E-O

O-E-O

O-E-O

O-E-O

O-E-O

O-E-O

O-E-O

O-E-O

Operational Issues

• Path computation tied to optical validation - not scalable

• Long protection / restoration paths not guaranteed

• Trouble shooting toolkit is limited ?

Wavelength Blocking

• Service provisioning needs same unused wavelength in all links

• 20-25% capacity stranding in medium sized networks (12-15 nodes)

Muxponder Inefficiency

• Stranded BW for sub- services (unless all services have same A-Z path)

• Problem gets worse for 40G line rate with mixed 1G/2.5G/10G services

All-optical ROADM is an automated patch panel - not a dynamically reconfigurable system

Muxponder Inefficiency:

Static Muxponders = Stranded Capital


Location C

10x10GbE 100G

Location A

Location B

2x10GbE

2x10GbE

1x10GbE

1x10GbE

Netw

ork

Span

Constrained Capacity

Remaining capacity locked into pre-determined A-Z endpoints

100G

10

0G

ROADMNetwork

Digital ROADMs offer flexibility & economies


Use (analog) photonics for what it does best: transmission

Use (digital) electronics for everything else – more functionality / lower cost

Digital Electronics& Software

Sub- add/dropSub- Grooming & switching Multi-Service MultiplexingProtectionSignal regenerationError correctionPerformance monitoring

Integrated

Ph

oto

nics

Integrated

Ph

oto

nics

optical (analog) electrical (digital) optical (analog)

Digital Optical Networking integrates ROADM


ROADM functionality built-in

Further enhanced with sub-grooming & switching

Automated configuration up to 8 degree’s

Integrated Colorless, Directionless & Contentionless

Eliminates multiple points of failure associated with cables & manual process

The simplicity of Digital ROADM’s enable the switching of optical paths form any point in the network, without fiber connections or expensive components.

Digital ROADM’s do not preclude optical express

Digital ROADM

OCG (multi-λ) Optical Express

Local add/dropAny service 155M - 100G

Digital Bandwidth Management

• Optical Express for cost-effective trunks

• Flexible Digital ROADM

Optical Express transmits groups of λs

Digital ROADM drops any service as needed

Provides ultimate flexibility and simplicity

Eliminates complexity of traditional ROADM

Lowering Cost & Complexity with Digital ROADM’s


93% fiber connections reduction

Eliminates costly WSS & AWG line modules

40-60% CAPEX reduction per ROADM node

Nationwide Network Modeling Summary


100

90

80

70

60

50

40

30

20

10Dep

loye

d C

apac

ity

(%)

Rev

enu

e G

ener

atin

g (%

)100

90

80

70

60

50

40

30

20

10

100GMuxponder

50%

40GMuxponder

66%

DigitalROADM

92%

• Large N. Am. Network Model: 33,084 route km, 47 core WDM links• About 10 Tb/s of customer service demands (network traffic volume)

Digital Optical Networks improve ROADMs


Digital ROADM’s offer maximum flexibility• Colorless – any wave to any color• Directionless – any wave to any degree• Contentionless – any wave re-use without contention

Digital ROADM enable management at the circuit level for maximum flexibility & efficiency• Pre-deploy affordable bandwidth blocks & relegate dynamic

reconfigurability to digital domain at the circuit level

To achieve optimum flexibility & efficiency, Digital Optical Networks are a proven best practice

Network Simplification & Cost Optimization


The Network Must Optimize Total Network Costs

• Scale net usable bandwidth• Maximize network ability to deliver BW services (ie, switching)• Minimize $/Gb (w/ switching)

• Increase Automation for scalability

• Optimize Space, Power• Minimize MTTR

ReduceOPEX

• Migrate packet transport to optical layer

• Converge DWDM & Packet/Circuit transport

OptimizeL0-L3 Cost

Scalable Optical

Infrastructure

Optimize TCO

Minimal End-to-end Cost

Optical Networking Evolution

WDM + OTN Convergence

Integration of: Scalable capacity (WDM) Bandwidth Management (OTN) Automation/Intelligence (GMPLS)

Converge WDM + switching• Integrated L0 – L1

Transport cost savings Digital “ease-of-use” Unconstrained bandwidth Bandwidth efficiency

DWDM

OTN

Packet

Integrated

WDM +

OTN +

GMPLS

Optical Networking Evolution

P-OTN Convergence

Integration of: Scalable capacity (WDM) Bandwidth Management (OTN) Automation/Intelligence (GMPLS) Packet Switching

Converge Packet + transport • Integrated L0 – L2.5

Total Network cost savings Further expand service simplicity

& network efficiency

Integrated

WDM +

OTN +

GMPLS

Packet

Integrated

Packet +

WDM +

OTN +

GMPLS

Digitally Switched WDM

Back-to-back Transponder Elimination

muxponder trib

End-to-end CapEx Streamlining Opportunities


Residual stranded bandwidth from constrained point-to-point delivery schemes must be considered

Integrated WDM + Switching creates E2E circuit cost reduction opportunities

Metro Core Subsea

$$$Conventional

ROADM

muxponderManual Back-to-back

MuxpndrManual Back-to-back

Muxpndr

$$$Conventional

ROADM

Swit

ch

muxponder

Swit

ch

xpnder

Eco

no

mic

Via

bili

ty

$$$

$

Automation for Enabling Dynamic Bandwidth


Controlling the network can no longer be a manual process

Auomationvia Intelligent Network Control Plane is essential

1. Automated network discovery

• Facilitates plug-and-play transport networking

• Reduces inconsistencies between network & OSS

2. Automated service delivery & bandwidth management

• Dynamic end-to-end circuit routing & provisioning

3. Dynamically (re)configurable transport capacity for CapEx & OpEx reduction opportunities

4. Integrated protection & restoration

Automation is not OPTIONAL, rather essential for OpEx & CapEx cost reductions


GMPLS Enables Scalable Network Automation

Topology Auto-Discovery

Restoration

Auto-Provisioning

Network automation Advanced services

• Network & resource discovery

• Dynamic service provisioning

• Automated Bandwidth accounting

• 50ms protection & GMPLS Restoration

• Dynamic bandwidth services (UNI based)

• Optical VPNs

Connection request• BW = 1Gb/s• Dest = Z

Dynamic Circuit Provisioning creates point-n-click process

GMPLS RSVP-TE enables robust dynamic circuit creation & deletion

• Automates allocation of capacity & assignment of crossconnects along circuit route

• Crossconnects correlated via circuit identifier

Robust signaling protocol employed for automated provisioning of Sub-network Connection (SNC)

• Handles failure modes & crank-back situations

• Circuit path information stored & retrievable at endpoint

• Native client signal PM monitored at SNC endpoints

LocalEndpoint

RemoteEndpoint

RESV RESV

PATH PATH

1. Compute best route

2. Initiate signaling request

3. Notify OS

GMPLS benefits


• Flexibly reconfigurable link capacity enables multiple circuits to time-share bandwidth

• Example: spot bandwidth for large data moves

Dynamic reallocation of digital capacity

• Scales with network growth, mitigates human errors

• Reduces Time to Revenue

• Improves service up-time with dynamic restoration

OpEx optimization through automation

• GMPLS UNI & open APIs for capacity delivery in realtime

• Adaptive to unexpected traffic changes or demands

High capacity bandwidth-on-demand

Summary

What I want you to take away today


Bandwidth elasticity is paramount to deliver services cost effectively wherever and whenever needed• Service-ready bandwidth is essential

Network flexibility must adapt in real-time to changing requirements, while optimizing efficiencies & minimizing cost

Automation is mandatory to ensure a long term sustainable & scalable OpEx model

Thank You

Chris Liou

[email protected]

enabling dynamic, rapid, and effective end-to-end bandwidth from metro edge to waters edge

Technology

latency bandwidth

subsea bandwidth

higher bandwidth

network elasticity

dynamic bandwidth enterprise

dynamic bandwidth allocation10

cloud network infrastructure

end bandwidth frommetro