ip-over-wdm integration strategies

© NOKIA Nasir Ghani, 2000NOKIA RESEARCH CENTER, BOSTON, USA

IP-Over-WDM Integration StrategiesIP-Over-WDM Integration Strategies

Nasir Nasir GhaniGhani, , PhPh.D..D.NokiaNokia Research Center Research Center

Boston, USABoston, USA

nasirnasir..ghanighani@@nokianokia.com.com

Sprint Research Symposium, University of Kansas, Kansas, March 8-9, 2000Sprint Research Symposium, University of Kansas, Kansas, March 8-9, 2000


• Introduction/Background

• Optical-Layering Approaches

• MPLS-Based Approaches

• Future Trends/Technologies

• Conclusions

Presentation OutlinePresentation Outline


• WDM enabling technologies- Fibers (SMF up to 600 km, dispersion optimization for more)- Lasers (2.5 Gb/s mature, wavelength programmability)- Amplifiers (wavelength/power equalization issues)- Increasing channel counts (C and L bands)

• Improving optical network elements (ONE)- Add-drop multiplexers (O-ADM), cross-connects (WRS/OXC)- Re-configurable operation, scalability/O-E challenges remain

• Networking applications- Multi-protocol support/transparency- Improved higher-layer connectivity- Traffic engineering/virtual topology control- Improved network survivability

Introduction/BackgroundIntroduction/Background


• Current deployment status- Many point-to-point, O-ADM/OXC in WAN now- Proprietary control, static provisioning of “circuits”

• Likely industry evolutions/migrations- Increased re-configurability (switching, λλ-conversion)- Improved survivability, traffic engineering (w. higher-layers)- New data framing solutions/formats - Futuristic: burst switching, limited/full packet switching- Address control plane issues (research, standardization)

• IP data traffic profiles- Over 90% is highly delay insensitive Computer-to-computer traffic, email, web, ftp- Highly asymetric profiles (time-of-day variations) Implies need for rapid reconfigurability- Multi-path diversity vs. single-path reliability

Introduction/BackgroundIntroduction/Background


• Overall features - New “circuit-provisioning access layer” concept Edge interworkings, increased costs- Multiple “client” protocols supported E.g., IP, ATM, SONET, frame relay, Ethernet

• New protocols required (optical UNI and NNI)- Optical provisioning protocols (RWA, survivability)- Higher-layer topology/resource engineering applications New, automated "inter-layer" protocols required - Limited transparency, i.e., new framing formats SONET/SDH, “SONET-lite”, digital wrappers- Standardization activities (ITU-T, OIF, T1X1, ODSI)

• Vendor offerings to date- Mainly proprietary (complete) solutions- E.g., Lucent, Nortel, various startups

Optical-Layering ApproachesOptical-Layering Approaches


Physical Layer

SONET

ATM

IP/MPLS

Optical Adaptation Layer

IP/ATM/SONET approach.IP/PPP into AAL5 over SONET,most framing overhead incurred,four management layers

Packet-over-SONET (POS)approach. IP/PPP/HDLC intoSONET framing (RFC 1619),three management layers.

IP-over-WDM approach.IP/PPP/HDLC packets directlyover optical lightpaths. Majorframing and fault-recoveryconcerns for optical adaptationlayer if SONET replaced. Stilltwo management layers.

Optical adaptation layer will manage WDMchannel setups/takedowns and provide some levelof protection/recovery. Additionally, this layermay likely introduce another framing function toreplace SONET-type framing.

Physical optical layer performs functions such asoptical amplification, wavelength switching andconversion, add/drop, O-E-O conversion, etc.

SONET

IP/MPLS

IP/MPLS SONET

Traditional trunkedSONET traffic(i.e., multiplexedvoice calls)



• Wavelength-channel provisioning - Routing and wavelength assignment (RWA) problem Maximize resource utilization, minimize costs- Various complications/constraints arise Analog concerns, λλ-conversion, control architectures- Addressing schemes/multi-protocol address resolution Extend ARP or use (adapt) NHRP-type solutions- Edge traffic aggregation issues (e.g., FEC, LSP stacking)

• Automated virtual topology control- Application driver for RWA algorithms- Improve network efficiency by “re-adjusting” topology Lightpath channel re-routing, λλ re-tuning, etc.- Complex “two-layer” interactions w. higher layers Operational timescales, information transfer issues - Standardized or proprietary solutions?




Higher-layer routing and

traffic-engineering protocols(i.e., virtual topology control)

Policy/operationsengineering

Data networktopology

User demands/traffic profiles

WDM RWA and survivability

provisioning algorithms(fast request resolution)

Offline network resource

optimization algorithms(longer-term, channel re-

routing/re-tuning)Lig

htpa

thre

ques

t/re

leas

e

Opticaltopology

Cha

nnel

sta

tus

Failureevents

Client Protocol Layer

Optical Protocol Layer

Resourceutilization

Resourceutilization

Analogconcerns

Joint higher-layer/optical-layerresource optimization algorithm

(offline, longer-term virtual topologyreconfiguration schemes)

Data flow routes(e.g., MPLS LSP,

ATM VC/VP)

Lightpath channel routes,wavelength assignments

Protocols interaction between optical and data layers


• Channel survivability schemes - Protection schemes via backup channels Dedicated/shared strategies can provide multiple levels- Restoration schemes also considered “Self-healing”, hard-guarantees difficult to pre-specify- Scalable fiber-level protection also possible Reduced signaling explosion for fiber-cut events

• Escalation strategy designs necessary- Many higher-layer protocols already provide recovery E.g., IP-rerouting, ATM protection rings, SONET/SDH APS- Destructive interference degrades responsiveness/efficiency “All layers do not switch over to same backup resource”- Escalation strategies for “coordinating” inter-layer recovery Top-down and/or bottom-up schemes proposed- Complex timing issues, topological considerations




Dedicated Backup

A

B

D

C

FE

G

Primarychannel A-D

Primarychannel A-E

Backup channelsA-D, A-E

Shared Backup

A

B

D

C

FE

G

Primarychannel A-D

Primarychannel A-E

Shared backupchannel hop(A-D, A-E)

Lightpath channel protection schemes


• Novel IP-based approaches for rapid provisioning - Re-use existing signaling framework (i.e., control plane) Less standardization, faster vendor interoperability - Direct “single-layer” integration (OXC/WRS ≅≅ LSR) “Optical lambda-switch routers” (O-LSR)- Abstracts lightpath to MPLS label switched path (LSP)- No addressing concerns arise (use IP addresses)

• Key MPLS features exploited- LSP tunneling (label stacking/swapping) - Explicit routing (ER) capabilities- LSP survivability capabilities- Constraint-based routing (CBR)/resource engineering

• Increasing industry momentum- Vendor proposals (Nokia, UUNET, Cisco, NTT) E.g., lambda-labeling, multi-protocol lambda switching- Standardization work ongoing (IETF, OIF)

MPLS-Based ApproachesMPLS-Based Approaches


Physical Layer

Frame Monitoring

IP/MPLS

IP-over-WDM with framing “sub-layer”for fault detection/localization purposes.IP MPLS is main provisioning layer,framing done in a strictly point-to-pointmanner between MPLS O-LSR nodes(e.g., SONET, “SONET-lite”, digitalwrappers, etc.).

Direct packets-over-lightpathapproach. For carrier-class reliability,this requires reliable, effective fault-detection/location and monitoring atthe optical layer, and tight-couplingwith the IP layer for highly-criticaluser services. Likely, this approachwill take a long time to emerge. Forless stringent requirements, moredirect framing w/o monitoring can beused (e.g., gigabit Ethernet)

IP/MPLS protocol will assume most of thefunctionality for data newtork and opticallayer provisioning. This includes lightpathsetup/takedown, protection/recovery, andpossibly even fault detection/localization.

Physical optical layer performs functions such asoptical amplification, wavelength switching andconversion, add/drop, O-E-O conversion, etc.

IP/MPLS

Framing “sub-layer” for fault detectionand localization functions only



• Lightpath LSP tunneling/routing- Subsume existing RWA algorithms Centralized, distributed, and hybrid architectures- Analogy between MPLS labels and WDM wavelengths No explicit label encapsulation/lookup required - "TTL pre-decrement" operation at edges I.e., use existing provisions for "MPLS-over-ATM”- Incorporate analog concerns (TLV/MIB definitions)

• Explicit-Routing (ER) functionality- Specify LSP routes, engineer resource allocation policies- Can subsume most advanced WDM protocols I.e., traffic, policy, priority, resilience, preemption attributes- Incorporate with IP/MPLS traffic engineering MPLS CBR (RWA) protocols, IGP (OSPF) updates- Improved provisioning, reduced operating complexity “Single-layer” facilitates topology/resource information flow- Note: Algorithmic complexities/concerns unchanged E.g., RWA,virtual top. control, traffic aggregation/mapping




I4

I6I1

I2

I3

IP network (edge not necessarily MPLS-capable)

Core MPLS O-LSR nodes withoptical switching fabrics

(performing “virtual” labelencapsulation and lookup)

Electronic MPLS LSR nodes withelectronic switching fabrics or

simply high-speed regular IP routers

I5

M7

M9

M8M10

M11

Optical “lambda-labeling”MPLS subnetwork

Edge MPLS O-LSR nodes performlabel merging/tunneling to/from higher-

bandwidth (coarser) “lambda” LSP’s

Large bandwidthlambda LSP’s,explicit routing

Small bandwidth LSP’s, hop-by-hop and/or explicit routing

label assignment withinelectronic MPLS cloud, or

simply regular IP routers w.LSP aggregation functionalityat edge MPLS O-LSR nodes.


Optical MPLS subnetwork(large, discrete LSP granularities)

L1

L3

L4L2

Ingress MPLS O-LSRnodes perform LSPstacking on to large

granularity lambda LSP’s

IP-MPLS network Egress MPLS O-LSRnodes perform lambda

LSP de-stacking to finergranularity LSP’s

M2

M1

Electronic MPLSLSR nodes handling

finer granularityflows (all label

operations)

Large (discrete) lambda LSP’sfor tunneling between edgeO-LSR nodes (optical labelswapping operations only)

M4

M3

M5

Optical MPLSO-LSR node

Regular MPLSLSR node

High-speed electronicMPLS LSR node,

capable of handlinglarge granularity LSP


Lambda (optical) LSP tunneling in WDM networks


MPLS Signaling(CR-LDP, RSVP)

Constraint-BasedRouting (CBR)

IGP Routing Protocols(OSPF, RIP, IS-IS)

Link-state databasew. extensions

RWA algorithms andvirtual topology controlschemes (i.e., joint withelectronic LSP control)

Optical topology/resourcedistribution, possibly optical

analog-related parameters(e.g., λλ-usages/conversion,

O-SNR estimates, etc).

Wavelength channelsetup and takedown,

protection/restorationswitchover coordination

Extensions for optical routingmetrics, resource engineering,

etc., re-use/extend existingIGP TLV/MIB definitions


MPLS Protocols Framework for Optical WDM Networks


• Lightpath LSP survivability- Joint provisioning of backup channels (w. RWA phase) Use ER function, switchover on fault detection- Edge-to-edge (path) and sub-path repair Generic protection switch/merge nodes, FIS- Restoration schemes also possible Active signaling/selective-flooding after fault- Label-stacking can incorporate fiber protection schemes- Reduced (no) multi-layer fault coordination concerns

• Fault detection and localization issues - Can still use electronic “frame-monitoring sub-layer” "SONET-like” timescales (e.g., SONET, digital wrappers, etc)- Optical monitoring for fault detection/localization Power-level monitoring schemes (not-standardized yet)- IP timer-based solutions for fault detection/localization Reduced keepalive/hello timers, subsecond restoration- “SONET-like” timescales not necessary for most IP traffic



M2

M1

M3 M4

M6

M5

M7

Optical MPLSNetwork

Primary workinglambda LSP

Backup (protection)sub-path route

Protection SwitchLSR (PSL) node

(O-LSR)

Protection MergeLSR (PML) node

(O-LSR)

M1-M8lambda LSP

source O-LSR

M1-M7lambda LSPsink O-LSR

Reverse faultnotification path


Proposal for MPLS LSP recovery


• Packet and burst switching evolutions- Reduced λλ provisioning timescales (ms to ns)- Large "container" packet switches exploiting WDM Electronic header processing overlap w. payload transfer, fiber loop/circulation buffers, optical label processing, etc.- Optical burst switching (OBS) concepts With/without (optical) buffering, ongoing research

• Hybrid network switching paradigms- Re-emergence of (optical) packet-switching in the core?- Hybrid nodal designs ("multiple features in single box") Combined circuit/packet switching- Trunking wavelengths across network (i.e., fiberpath) Enabler: MEMS technology, driver: increasing penetration- Fiber-wavelength-packet (FWP) node

• Optical-layering vs. MPLS-based approaches- Packet/burst/hybrid switching easily incorporated w. MPLS- Optical-layering requires added extensions/modifications Likely overlap/redundancy w. many IP features

Future Trends/TechnologiesFuture Trends/Technologies


Hybrid designs:optical packet/labelswitching w. circuit-

switched wavelength/fiber switching layers

Present 2-5 years 5-10 years

Direct optical labelprocessing and

packet buffering

Adaptivecircuit

switching(sec-ms)

Staticcircuit

switching(days)

Coarsepacket

switching(ms)

Timeline

Sw

itch

ing

Tec

hn

olog

y ( λλ

pro

visi

onin

g ti

mes

cale

s)

10+ years

Finepacket

switching (ns)

Adaptive wavelengthrouting, increased nodal

reconfigurability(active signaling)

Large “container”packet switching

(wavelength-aware)

Static wavelength routing,limited reconfigurability

Optical burstswitching (OBS)

node designs




Wavelength switching/conversion matrix

Incomingfibres

Outgoingfibres

Demux Mux

WDM dropchannels

WDM addchannels

Electronic switching matrix

Inputbuffering/processing

Outputbuffering/processing

Packet Switching Level

Wavelength Switching Level

Fiber Switching Level

Multi-fiber spatial switch

Fiber drop Fiber add

Input datalinks

Output datalinks

O-E

WDM inputs(from laserinterfaces on

routers)

WDM outputs(to laser

interaces onrouters)

E-O

Hybrid Fiber-Wavelength-Packet (FWP) Node


ConclusionsConclusions• WDM networking growth strong, new provisioning paradigms

- Scalability, falling costs, increasing penetration- Wavelength switching timescales will decrease Weeks ⇒⇒ days ⇒⇒ hrs ⇒⇒ min ⇒⇒ sec ⇒⇒ ms ⇒⇒ ns (?)

• Optical-layering approaches- Define (standardize) a new, separate optical layer: Provisioning protocols, addressing, interworking protocols- Pros: Multi-protocol transparency, basic solutions exist- Cons: Added interworking complexity, standardization issues (delays), current/future migration concerns

• MPLS-based approaches - “Re-use” MPLS provisioning/control plane framework: Subsume many advanced concepts in IP domain- Pros: Expedited interoperability, easier IP integration, smoother future migration- Cons: Entrenched mindsets, current investments, concerns with “opening up” optical core


ReferencesReferences

• N. Ghani, “Integration Strategies for IP Over WDM,” Optical Networking Workshop(co-sponsored by IEEE Region 5, ACM, SPIE) Dallas, TX, January 2000.

• N. Ghani, “Lambda-Labeling: A Framework for IP Over WDM Using MPLS,” toappear in Optical Networks Magazine, April 2000.

• D. Awduche, et al, “Multi-Protocol Lambda Switching: Combining MPLS TrafficEngineering Control With Optical Crossconnects,” IETF Draft, draft-awduche-mpls-te-optical-01.txt, November 1999.

• N. Ghani, “Survivability Provisioning in Optical MPLS Networks,” to appear in 5th

European Conference on Networks & Optical Communications, Stuttgart, Germany,June 2000.

• K. Kompella, et al, “Extensions to IS-IS/OSPF and RSVP in Support ofMPL(ambda)S,” IETF Draft, draft-kompella-mpls-optical-00.txt, February 2000.

• N. Ghani, et al, “On IP Over WDM Integration,” IEEE Communications Magazine,March 2000.

• N. Ghani, S. Dixit, “Channel Provisioning for Higher-Layer Protocols in WDMNetworks,” Proceedings of SPIE All Optical Networking Conference: Architecture,Control, and Management Issues, Boston, MA, September 1999.

• H. Chaskar, S. Verma, R. Rayadurgam, “A Framework to Support IP Over WDMUsing Optical Burst Switching,” Optical Networking Workshop (co-sponsored byIEEE Region 5, ACM, SPIE) Dallas, TX, January 2000.

ip-over-wdm integration strategies

Documents