Optical Networking Overview of the Terminology Technology Architectural

Optical Networking Overview of the Terminology, Technology, Architectural Issues and Aspects of Associated Business/Financial Issues Joint Techs/APAN Conference Honolulu, Hawaii January 29, 2004 Jerry Sobieski 1 Director, Research & Technology Development Mid-Atlantic Crossroads

Disclaimers n n This presentation is only meant as skeleton for discussion The speaker is not the expert in all (or necessarily any) of the topics discussed

The Justification for R&E customized optical networks n The commercial sector has trouble convincing itself to build leading edge networks unless/until there is enough uptake to generate [sufficient] return on the investment The R&E community isn’t large enough (by itself) to justify the investment by the carriers n The R&E community have rather different requirements than the business/commercial sector n The R&E community can’t afford commercial pricing n

Why does the R&E community think they are able to build them? n The resources are provided at a “cost” basis rather than a “market” basis n n I. e. the profit motive is [in theory] substantially less, hopefully reducing TCO The R&E community collaborates, where as the private sector competes – shared costs and combined financial resources The “early adopter” service requirements are different than commercialized production services The R&E community must build them in order to have the infrastructure required to experiment and develop new service models and capabilities

Why is “Optical” Networking so important? n n n Understanding -To explore, we need a common basic understanding of the optical technology, capabilities, architectural/engineering tradeoffs, and future evolutionary roadmap Experience - We need real examples and first hand experience to truly understand the implications for the R&E needs. Prioritization - We need to identify and prioritize our requirements and map them to our resources (I. e. finances, partners, facilities, timeframes, etc)

Outline of Discussion n Brief review of optical concepts and current technology Discussion of architectural design considerations when planning an optical deployment Discussion of some of the business issues associated with optical services

What Optical Network? n n n Most folks think of wave division multiplexing and all fiber (no copper) based connectivity A few think of the “all optical” or “passive” or transparent optical transport properties of optical networking The cognicenti include all optical switching, buffering, and other packet processing.

Optical Networking Building Blocks n Network elements (nodes) n n Laser types, characteristics, coding formats Transponders, transceivers Multiplexors and demultiplexors Switching technologies n n n Protection switching OADMs Wavelength switching Amplification and regeneration Fiber (links) n n n Fiber types, characteristics Applications (campus, metro, long haul, etc) Engineering issues n n Attenuation, Dispersion Other non-linear effects

Network Nodes n Optical nodes need to combine several functions: n n n Add/drop of wavelengths Wavelength conversion to CPE interfaces Signal regeneration Wavelength routing (switching and/or translation) Wavelength amplification/equalization

Components and Terminology n Optical Add/Drop Mux (OADM) n n n Adds or drops a wavelength(s) to/from the fiber, Passes (ignores) other wavelengths. Optical channel modules n n Convert traditional laser interfaces to ITU “grid compliant” wavelengths (e. g. 1310 nm to/from ITU 33) Perform other regeneration functions (retiming, reshaping)

Optical Add/Drop Multiplexor n n Two fiber example Possibly from a ring configuration Mux Dmux OADM Dmux Mux Channel Modules

Components and Terminology n n Wavelength Router/Switch n Routes wavelengths presented at an inbound port to some specified outbound port. n Wavelength routers do not change the actual wavelength – just separate and recombine wavelengths between fiber ports n Some operate on each wavelength individually, some operate on wave bands (groups of contiguous ITU wavelengths) Wavelength Conversion n Convert one wavelength to another n Typically requiring an intermediate electrical step (OEO)

Components and Terminology n Wavelength Translation n Copy the modulated signal from one wavelength to another n All Optical n Amplifiers n All optical device used to amplify optical signals n Erbium Doped Fiber Amplifier (EDFA) n n Utilizes a “pumping” laser and erbium doped fiber to amplify all signals in a broad range of wavelengths (C band lambda’s ~1450 nm to 1600 nm) Raman amps – better SNR, higher power optics

Current technology issues/limitations n Transponders use fixed ITU wavelengths n n n Requires different hardware to provision different wavelengths Tuneable wavelength lasers still prohibitively expensive Demultiplexing still requires fixed hardware for specific wavelengths (tuneable filters? )

Optical Architectures n Design Objectives n n Advanced Technology? Cost reduction? Both? Cost efficiency – reduce the overall cost of providing necessary services n What are the “necessary” services? n n n Production needs of the users – I. e. dependable, inexpensive, but [typically] not technologically leading edge. Ex: Commodity internet access Advanced (“experimental”) capabilities for research applications – I. e. new types of services that support emerging applications requirements. Network research and experimentation

The Overlay Model n n n Upper layers (e. g. layer 2/3) are unaware of the underlying transport layer Simple (from the upper layers) – consistent with current sonet transport layers Upper layers have no control or knowledge of lower layer topology

The Peer Model n n Network layers interact with the transport layer to request resources Implies a control plane interface (and some level of routing interaction)

One Potential No. F: Concept A IP Router Fiber Routes X X X International links X OADM X X X RON or campus Wavelength X Router

Characteristics of Concept A n n National administrative domain Optical transport layer – peer model n n Fewer IP routers – but probably bigger (!) n n n Dynamic & flexible bandwidth provisioning Full mesh between core routers using diverse lambda routing Delivers IP/lambda to gigapops/RONs Gigapops can peer at optical layer and/or IP layer. n IP peering at multiple core nodes

Current limiting factors in optical technology and deployment n Tuneable (sp? ) wavelength lasers and/or filters n n Optical switches still not mature n n Required for efficient wavelength routing New, unproven, and still evolving Interoperability between vendors is. . better n n n Control plane standards are maturing (e. g. GMPLS) Management protocols and implementations vary widely Important protocol issues have not been resolved or standardized for inter-domain operations

Limiting factors (cont. ) n Integration of WDM technology directly into routers, layer 2 switches, workstations, etc. n Interfaces that operate on the ITU grid will eliminate OEO conversion at the WDM interface n n n “ITU gbic”s – eliminate OEO stage Optical UNI Access to dim/dark fiber is still non-trivial and expensive task n n In the major metro areas it is improving But the rural land grant institutions are still struggling n Many Universities are not topologically near major telecom hubs or in dense, fiber rich metro regions. How do we solve this problem?

Optical Network Design Objectives n Cost Efficiency and Advanced Technology are not diametrically opposed concepts n Regional optical networks provide a significant flexibility to the user community: n Service capabilities are defined to the community’s needs n n Do not require the critical mass business case typical of large commercial carriers (Note: this does not mean these services can be provided for free!) Multi-institutional involvement allow for effective utilization of the investment n n making large investments in regional infrastructure possible in the first place, and reducing the individualized costs of services to each institution

Optical Network Design Strategies: Choosing the Points of Presence n Careful selection of peering points and/or Po. Ps n Locations that will provide the necessary interconnections and services for the long term – 7 years or longer (upstream services) n n Lit services Fiber access Provider competition (two friends are better than one ) Establish Po. Ps that benefit otherwise telecom challenged neighboring regions/metro areas (downstream services) n Increases consortial critical mass

Optical Network Design Strategies: Choosing the Points of Presence n Long term prospects and value of the network POPs allows for long term investment by the network and served community in fiber to one (or more) of the POPs n n Universities are stable (to a fault ) Commercial “telco hotels” provide [relatively] easy cross connects to/from the RON and other service providers n n Colocation space availability – consider expansion requirements over the long term, personnel access issues, etc Vendor neutral Meet-Me rooms National and international telecom access Able to incorporate private fiber built in from user community n Entrance/access permission for fiber provider

Points of Presence A MAX Example CLPK DCGW College Park, MD University of Md -Verizon, ATT, Qwest, Fibergate, Yipes -NGIX -NASA, NLM/NIH, NOAA, USM DCNE Washington, DC North. East George Washington University Qwest Communications Terapop - Verizon, Qwest, Level 3, RCN… -Qwest (primarily), MFN, Verizon, others -GWU, Georgetown, -Abilene, Esnet, Qwest DIA, Bossnet, … ARLG CUA, USNO … Arlington, VA USC/ Information Sciences Institute East - Verizon, Qwest, MFN, Level 3… -ISI-East, NSF, NCSA Access, …

Optical Design Objectives Interconnecting the POPs n Fiber architecture needs to be: n Of suitable grade and/or quantity to carry anticipated services n n n Diverse whenever possible to address redundancy and survivability issues n n n DWDM capability and modulation rate are a primary concern for current and future transport services Often simply lighting new fiber pairs is more cost effective (and sometime the only viable way to support new technology rollouts. ) Rings are the traditional method, meshes are becoming more common Minimize path length to reduce cost and span-length engineering complexities Able to incorporate private fiber built in from user community n Entrance/access permission for fiber provider

Initial Fiber Plans A MAX Example

JHU BALT Fiber Routes NARA NLM NIH GU NOAA HHMI USNO CLPK USM GSFC UMCP DCGW DCNE NWVA GWAS GWU ARLG PAIX MCLN GMUA ISIE NCSA

Optical Network Service Objectives n Balancing act - Anticipating long term optronics and fiber engineering requirements and dependencies is difficult at best n n MUST consider the useful life of current technology and plan for rollover However, cannot wait to deploy current advanced technology until futures are resolved – you won’t make progress. n n IMO, three year technology cycles is reasonable – useful life may be longer for less The types of services the RON will provide near term will drive fiber and optronics requirements: n n Anticipated transmission rates require careful attention to dispersion characteristics of the fiber, span lengths, ILAs, etc Types of transport to be provided will require careful selection of optronics components, wave plan, etc (e. g. TDM vs WDM vs SDM)

Optical Network Service Objectives n Range and dynamics of the optical network will increase complexity n n Persistent point to point transport in the campus/metro area is simple(r) Protected rings, meshes, aggregated services are less simple Long haul transport systems are not simple Dynamically reconfigurable and/or shared dedicated services are complex

Optical Network Service Objectives n Interactions between the transport layer and higher layer services must be understood n n E. g. automated protection switching implemented at the optical layer, framing layer, and/or IP layer can lead to protection storms Control plane implications must be considered for Peer model n n E. g. Is the control plane for the transport layer carried within upper layers? What happens to the control plane when/if a transport layer failure occurs? Ctrl Plane issues may be important if only for operations and management of the network (I. e. even if user access is not allowed)

Business Management Issues of Optical Services n Service Definition n What does the user actually receive? n n What are the service guarantees? n n Sonet? Ethernet? Fibre. Channel? ITU wavelength? Where is the demarc? (a centralized POP or user prem? ) Is this persistent? Dedicated? or shared? (I. e. TDM vs Statistically allocated) MTBF - Protected? Two 9’s or five 9’s? MTTR? What is the term of the contract? n How long will you be committed to providing this service? n n n Longer terms provide better amortization rates (to the user) Longer terms may increase risk to provider: out year costs may not be known Shorter terms are more front loaded capital intensive

Pricing Optical Transport Services n Cost basis: n n Cost = ammortized costs +incremental +operations+ depreciation + margin Ammortized infrastructural costs n Capital expenditure for base infrastructure (first operational wave) n n n Fiber IRUs, construction/improvements, network elements, amplfiers, etc along path Cost of money… Incremental cost of additional waves n n End points plus intermediate components (amps, mux/dmux, regen, protection components, etc. ) Incremental costs are [in general] difficult to generalize to a fixed cost/wave n Different end points will require different intermediate components (e. g. amps or regen), and the costs associated with specific paths may vary

Pricing Optical Transport Services n Cost Basis (cont. ) n Operating Expenses – monitoring, maintenance, provisioning, etc n n Function of network activity – more changes create higher operations costs Includes other non-capital expenses such as colo lease, office space, insurance, etc. Sparing is currently an expensive prospect with hardware specific transponders Personnel requirements: skill sets, coverage and availability, etc n n n Optical (WDM) engineering experience is scarce Support equipment (test lab, field test gear- OTDR, OSA, BERT, etc) Depreciation - Will you have any value left in the optical system when it is time to upgrade? n When the current generation is obsolete and has no remaining value, how will you finance the next generation?

Pricing Optical Transport Services n Service Pricing n n Price = costs + margin Ammortization and Depreciation costs are function of uptake rate and obsolesence rate n n n How many waves will be provisioned and paid for initially, or over the life of the optical system? What is the expected Lifespan? The more uptake, the lower the amortized overhead per wave – more affordability to the user community Margins (or the “P” word: Profit). Even not-for-profit organizations need to see margins on certain activities n n These funds enable new upstream activities (R&D) and can be used to cover unexpected expenses (e. g. relocation of a manhole for road expansion…) Margins provide operating financial buffer to address cash flow issues associated with service provisioning, billing, payment process.

A Brief Sampling of Financial Considerations n Even R&E not-for-profit network initiatives look a lot like a small startup enterprise: n Up front capital is required n n n Cash flow (not just annual budgets) must be addressed n n Loans, Investments, Grants Leasing rather than purchase of equipment can reduce some capital outlay Delays in fee payments can be devastating to a small organization Operating capital is crucial to buffer against payment jitter A business model/plan needs to be in place to cover operational expenses and recover the investment over time Service tailoring – e. g. not all users need dedicated optical services n Some institutions balk at paying overhead for infrastructure only a few [or other] institutions will use

Mahalo