TCOM 513 Optical Communications Networks

1
TCOM 513Optical Communications Networks

• Spring, 2006
• Thomas B. Fowler, Sc.D.
• Senior Principal Engineer
• Mitretek Systems

2
Topics for TCOM 513

• Week 1 Wave Division Multiplexing
• Week 2 Opto-electronic networks
• Week 3 Fiber optic system design
• Week 4 MPLS and Quality of Service
• Week 5 Heavy tails, Optical control planes
• Week 6 The business of optical networking
  economics and finance
• Week 7 Future directions in optical networking

3
Where we are

• So far, weve examined
• How optical networks operate
• How to put them together
• Important questions remains
• What kind of network should I build?
• Size/capacity
• Where should I concentrate?
• Core
• Metro area
• Local access

4
Topics for tonight

• Parameters affecting optical networking
• Pricing of network-type commodities
• The networking industry after the Big Crash
• Convergence
• Opportunities

5
Uses of optical technology

• Campus-short haul
• May need to trench fiber, build network yourself
• Connecting buildings across town
• Trenching fiber not generally practical
• May want to lease dark fiber or wavelength
• Going long distances
• Unless you want to become IXC, forget about
  laying your own fiber
• Use fiber/service already in ground
• Dark fiber
• Amplifiers, repeaters, etc. provisioned by
  service provider
• You have to do rest of engineering
• Wavelength service
• You have to modulate/demodulate
• SONET
• You just connect your equipment to a standard
  interface box

6
Parameters affecting optical networking

• Major players
• User/consumer
• Internet service provider (ISP)
• Telco/CLEC
• Interexchange carrier (IXC)
• Long haul pipe provider
• Supplier
• Builder/installer

7
Players (continued)

• Other interested parties
• Regulators
• Legislators/policy makers
• Consumer groups
• Environmentalists

8
Issues

• Supply side
• How to finance construction
• How to estimate return on investment (ROI)
• How to make a business case for investment
• How to ward off obsolescence
• Demand side
• How to price to stimulate demand
• How to handle network externalities
• Network externality price people are willing to
  pay depends on how many other people are using
  network
• How to allocate costs to equipment
• How to price shared services
• What should be free?

9
Issues (continued)

• Regulatory side
• What should be subsidized, and by whom?
• What rules should be in place?
• Who should be protected?
• How to ensure that time constants for regulatory
  process are shorter than those for technology
  being regulated

10
Supply and demand Economics 101
Highest price for full Network utilization

Demand per vendor (diluted by competition)
100
Revenue
Utilization
High
Medium
Low
Price
Price
11
Bandwidth growth (demand) models
Actual
Assumed
12
Limitations to bandwidth growth

• Demand slows due to legal issues
• Generally centered around IPR
• Digital distribution means perfect copies
• Internet radio tax
• Users can only absorb change at a certain rate
• Change usually means upgrades
• Forklift upgrades a hard sell to individuals and
  businesses
• Many are comfortable with current paradigm
• Regulatory questions are often vexing
• Universal service
• Cross-subsidization
• Business case for massive rewiring of U.S.
  difficult to make
• Uncertainty in demand
• Payback period may be extended

13
Brief history

• 1980s RD phase of optical networks
• Early 1990s Major deployments
• Late 1990s Dot-com, venture capital boom
• Massive build-out
• Business case? Who cares? Money spigot is on
  full!
• Set up a company, develop some product, sell out
  to Cisco, Lucent, or Nortel, or have an IPO, then
  retire rich!
• 2000-2002 The big crash
• Build out causes oversupply
• Optimistic demand forecasts do not materialize
• Prices drop
• Huge debt load cannot be serviced
• Next stop Chapter 11
• 2003-2005 Slow recovery, legal and regulatory
  battles

14
Pricing of network services

• Overview of network services
• Guaranteed, elastic, traffic contracts, network
  control
• Guaranteed and effective bandwidths
• Definition of effective bandwidths
• Relevance for pricing
• Economic models for networks
• Basic
• Finite resource sharing models/congestion pricing
• Implementation mechanisms
• Pricing for elastic services
• Pricing for guaranteed services

15
Some thoughts on network pricing

• There is no single view on charging for network
  services
• Disparate models, contradictory proposals
• There is no need to price (or charge for) network
  services !
• No congestion in the future !
• To cheap to meter
• Charge only for content
• There is nothing new in network pricing
• Economists have said everything before
• Never mind that they havent agreed
• Need abstraction, model to work with

Network Abstraction
Economic Theory
Model
Source Costas Courcoubetis/U Crete
16
Why charge for network services?

• Charging is not only for making profits
• Improving network performance
• Providing stability and robustness
• Creating revenue (to pay debt service)
• Charging should provide
• Important information for network control
• Information for buildout of network
• Pricing should have certain characteristics
• Simple but not simplistic
• Understandable
• Implementable
• Competitive

17
Network services as sold to users

• Packet or cell transport
• Contract usually has two aspects
• Connection or flow performance to be provided by
  network
• Traffic profile to which user must conform

Network
Service Interface
18
Network service types

• Guaranteed
• Network provides some form of performance
  guarantees
• Loss
• Delay
• Jitter
• User may request some amount of network resources
• May be subject to admission control
• Typical of ATM, FR networks with Committed
  Information Rate (CIR)
• Commonly needed for isochronous traffic (e.g.,
  video)

19
Network service types (continued)

• Elastic or best effort
• No specific performance guarantees
• Performance deteriorates during overload periods
• No specific bandwidth request-users can use all
  available bandwidth
• Intended for applications that can adapt sending
  rate

20
Guaranteed services

• Performance guarantees
• Quality of service (QoS) loss, delay, jitter,
  BER
• May be statistical (e.g., loss lt 10-7) or
  deterministic (delay lt 25 ms)
• Required mechanisms
• Connection admission control (CAC)
• Policing
• User-network traffic contract
• Provides connections QoS and traffic description
• Network promises to support specified QoS
  provided that users traffic satisfies his
  specified profile

21
Elastic or best effort services

• No specific performance guarantees
• Provide some form of fair treatment to different
  connections
• Feedback mechanisms used to inform source of
  congestion or other problems
• Explicit (e.g., data rate)
• Implicit (e.g., packet loss)
• Mechanisms programmed into routers, switches to
  share bandwidth, enforce fairness
• Source behavior
• Increase (additive) when no congestion
• Decrease (multiplicative) when congestion present
• Example TCP flow control

22
ATM Forum Service Categories

23
Enhancement of IP Infrastructure to Support
Diverse Set of Applications

• Service providers and network managers operating
  multiple networks to support range of
  applications
• This is not desirable from economic and
  maintenance standpoint
• IP infrastructure devices becoming cheaper due to
  proliferation of the public Internet and private
  networks
• Routers/switches and transmission
• Current IP infrastructure needs enhancement to
  support voice, video, and data at acceptable
  levels
• Flow of real-time bit streams

This is the challenge for the decade
24
VoIP Versus Conventional Telephony
Conventional Telephony Dedicated End-to-End
Connection
Tandem Switch
Central Office
Central Office
IP PBX
IP Wide Area Network (WAN_ (e.g., Internet
ISP
LAN
VoIP No End-to-End Dedicated Connection
Packets Take Best Available Path Through Network
25
Benefits Claimed for VoIP Networks

• Cost savings
• Simplified, more rapid provisioning
• Easier management
• Less maintenance
• More rapid deployment of new services
• Unified messaging

26
Risks Associated with VoIP

• Regulatory uncertainty
• Immature protocols and standards
• Data, telecommunications clash of cultures
• Unproven reliability and availability
• Security issues with IP networks
• Potential need for major network upgrades
• Enhanced 911 problems
• Network ubiquity problems
• Immature billing

27
Generic VoIP Architecture (Local and Long
Distance)

IP PBX
IP PBX
IP PBX
IP PBX
Medium-Large Site with IP PBX
28
Real-Time / Multimedia Requirements

• Support for a range of diverse applications
• Support for a range of bandwidth
• E.g., 128 Kbps collaborative video
  conferencing to 45 Mbps video-on-
  demand
• Support for a range of performance for voice,
  video, multimedia, critical data
• Delay, delay variation, packet loss
• Support a range of communication models
• Point-to-point, multipoint, multicast, broadcast
• Use of QoS for cybersecurity looks promising

29
Solution Alternatives

• Massive overbuild
• Brute force approach
• Feasible in good old POTS days
• Due to fractal nature of Internet traffic,
  difficult to know how much capacity is enough
• Fractal self-similar on multiple time scales
• Quality of Service (QoS) / Class of Service (CoS)
• Preferentially routes packets based on type of
  traffic they carry
• Does require software and / or hardware upgrades
• Complex nature of Internet and other networks
  makes prediction of performance difficult

30
Time scales of network control
31
Network control and pricing

• Set of feasible service offerings depends on
  network control mechanisms and provisioning
  capabilities
• Economic incentives influence both of these
• Network control
• Controls cell or packet flows to guarantee
  contracts
• Pricing
• Controls demand in order to improve efficiency

demand
32
Network control and pricing (continued)

• Prices differentiate quality of service, not
  content
• Prices depend on demand
• Driven toward cost by competition
• But set on margin
• Prices proportional to
• Effective bandwidth for guaranteed services
• Throughput for best effort services

33
Economic models for networks

• Context
• Communications services are economic commodities
• Exist within law of supply and demand
• Supply
• Amount produced or available
• Determined by technology of network elements,
  including management and control, cost
• Demand
• Amount users want
• Trade-off between QoS and willingness to pay
• Affected by competition, business climate
• Market interaction between service providers and
  users comes through price

34
Economic models (continued)

• Terminology
• Price associated with unit of service
• Tariff price structure as function of demand
• Common general form a bx, where x is demand
• Instrument for pursuing policy objectives
• Charge amount to be paid
• For more details, see
• http//www.aueb.gr/users/courcou/presentations/it
  c99/sld001.htm

35
Economics in telecommunications

• Large amount of work done on design of public
  utility tariffs for telephony
• 1970s peak load pricing, cross-subsidization
  (local by long distance)
• 1980s, 90s cost-based (subsidy-free) pricing,
  nonlinear pricing, peak load pricing under
  uncertainty, others
• Does not directly translate into good models for
  data
• Sharing of equipment/facilities
• Burstiness of traffic
• Complicated by diversity of new user-network
  contracts, multiplexing capabilities, resource
  sharing models

36
Sharing finite resources

• Network resource management occurs in various
  time scales
• Short time scales
• Amount of resources is fixed
• Control deals with optimal sharing
• Long time scales
• Resources expanded in order to improve average
  performance
• Marginal cost pricing combines above time scales,
  so operates in time frame of output expansion
• Short time scale problem
• Prices used to control way resources shared
• Can be used as input for deciding capacity
  expansion

37
Congestion pricing

• Types of congestion
• Demand exceeds capacity (I)
• Overall performance depends on usage patterns
  (II)
• Technology suggests that (II) is of limited
  interest
• Under heavy multiplexing, solution of (II) -gt
  solution of (I)
• (I) is relatively simple (add more capacity or
  raise prices)
• (II) requires performance model of network

usage
38
Congestion pricing summary

• Congestion price p controls sharing of C
• Determination of p based on network measurements
• Users i and j solve local optimization problems
• Need only prices posted by network
• Adjust x (network utilization)

39
Expanding capacity

• Benefit B of operating network link of capacity C
  under congestion price p
• r cost (fixed or variable)
• To maximize B, solve dB/dC 0
• If p gt rvar then buy more capacity until p rvar

40
Pricing at different time scales

• Congestion occurs at different times of day
• Implies hierarchy of decisions for users
• Higher level how much to transmit during day,
  night
• Lower level How fast, priority, when

41
Pricing at different time scales (continued)

• Lower level decision depends on more detailed
  information
• Time-of-day pricing typically takes into account
  average congestion during that period, qnetwork
  utilization

42
Implementation approaches
43
Charging for best effort services

• Elastic demand
• Flexible contract with network
• No guarantees on delay, throughput
• Examples TCP/IP, ATM ABR, UBR
• Sources of randomness
• Number of users
• Amount of data
• Amount of available resources

44
Charging for best effort services (continued)

• Need for flow control to regulate traffic
• Notions of fairness tend not to be economically
  efficient
• Goal provide optimal economic sharing of
  resources
• Two time scales
• Fast flow control maintains orderly flows
• Slow adjustment of prices regulates demand
• Common approach congestion pricing
• To reduce excess pricing
• To account for delay costs
• Prices can be computed two ways
• Dynamically based on demand
• Approximated from historical (time-of-day) demand

45
Charging for best effort services (continued)

• Issues
• Cost of computing prices
• Cost of exchanging information with users
• Stability of pricing mechanism
• User preference

46
The telecom world today after the Big Crash

• Overview of the current situation
• Evolution vs. revolution
• The growth paradigm
• The culture clash
• The winning strategy
• Fear vs. Greed
• Future architecture

47
Current landscape

Content providers
Specialized (horizontal) Service providers
Vertically Integrated providers
Access Internet Long-haul Application
Optical equipment manufacturers
Optical fiber manufacturers
Legacy equipment manufacturers
48
Major players

• Optical equipment manufacturers
• Nortel
• Lucent
• Alcatel
• Ciena
• Cisco
• Juniper
• Sycamore
• JDS Uniphase
• Optical fiber manufacturers
• Corning
• Pirelli
• Alcatel

49
Major players (continued)

• Optical network providers
• Major IXCs (ATT, Worldcom, Sprint)
• Level3
• Global Crossing
• Content providers
• AOL
• Internet Portals (e.g., Yahoo, Google)

50
How to think about the networking problem
Source J. McQuillan/NGN2000
51
How to think about the networking problem
(continued)

• Evolution
• Future looks like the present, only more of it
• Same players, evolving technology, increasing
  demand
• Revolution
• Next Generation (NG) providers will build NG
  networks using NG technology from NG vendors
  using NG funds from (??)
• New applications, new ways of doing things, new
  paradigms will chase out the old
• Organic growth
• Old and new will coexist in more complex, larger
  whole

52
Paradigm shift crash of the 1990s model
Source J. McQuillan/NGN2000
53
Old paradigm

• Grow through acquisitions
• Use increasing value of your stock to finance
  purchase of startups or competitors
• Create positive feedback loop
• Commonly used by major players
• Cisco
• Lucent
• Nortel

54
Old paradigm (continued)
Source J. McQuillan/NGN99
55
Old paradigm (continued)
Source J. McQuillan/NGN2000
56
How the world changes
Source J. McQuillan/NGN2000
57
New paradigm

• Not yet emerged
• Likely to focus more on value than utility
• Wild predictions wont justify investment
• Jupiter By 2005, 75 of retail sales will be
  online
• Too many people burned in dot com flameout

58
The culture clash
Source J. McQuillan/NGN2000
59
Culture clash (continued)

• Bell heads
• Name sometimes given to RBOC workers
• Refers to anyone who thinks in terms of
  enterprise networking
• High reliability and availability
• Relatively slow pace of change
• Priority on customer service
• Often thought of as voice oriented
• Net heads
• Name given to IT folks who value network
  connections
• Different set of values
• Innovation outweighs inconveniences of crashes
• Customer service a low priority
• Often thought of as data oriented

60
Culture clash (continued)

• Bell heads We know how to build and operate
  massive mission critical networks, so we will
  control the Next Generation networks
• Net heads We are the only ones who get it, so
  we will control them

61
Who will be the winner?
RBOCs
CLECs, GXCs
Source J. McQuillan/NGN2000
62
What will be the winning strategy?

• Vertical integration
• Traditional telephone/telecommunications model
• Own all facilities and data centers
• Offer customers one-stop shopping
• Horizontal specialization
• Offer best in class services
• Lease facilities/services from others
• Do one thing well
• Take advantage of standard interfaces

63
Clash between fear and greed

• Fear profits are most important
• People invest money expecting that income will
  exceed expenditures for any company
• Value investing
• Greed only growth matters
• Grab market share and increase volume
• Dont worry about anything else
• Utility investing make a useful product,
  everything else will take care of itself

64
Fear vs. Greed
Greed Era
Fear Era
P/E
SP 500 Index / 10 year moving average of profits
Source Robert Schiller/Yale J. McQuillan/NGN2000
65
Future architecture

• Innovation
• Multi-service switches will allow support of all
  traffic
• Legacy
• Internet as a service
• Services will otherwise remain separate
• Very efficient at what they do
• Convergence
• Internet (or private IP network) will be bearer
  of all services
• Voice
• Data
• Video

66
Convergence Topics

• Definition of convergence
• Types of convergence
• History of convergence efforts
• Drivers (benefits)
• Impediments
• Rollout

67
What is convergence?

• Convergence is a buzz word often used
• Brandished as wave of future
• Meaning far from clear
• Questions of interest
• To what extent is convergence real and wave of
  future
• To what extent is convergence just hype
• Basic idea use of some common infrastructure to
  support multiple services

68
Levels of convergence
Telecom/IT (e.g., ASPs, network computing)

Content Business
reflects
Application (e.g. music, video over Internet)
Increasing visibility to usercloser to edge
Switching (e.g., VoIP)
Transport (e.g., multiplexed channels)
69
Why convergence?

• Most significant is a vision of the future a
  more elegant world where everything is simpler
• Most telecom services are a result technological
  limitations existing at the time of their
  introduction
• Bandwidth of early telephone gear (200-3000 Hz)
• DSn bandwidth is multiple of 64K used for voice
• NTSC color TV state-of-art in 1940s
• In theory, convergence promises a clean slate
  approach
• Everything (or at least many things) reengineered
  to provide better, more flexible service to the
  user
• In practice, a mixed bag, with some successes,
  many disappointments, areas still under
  development

70
The present situation

• Current telecom infrastructure worldwide is
  circuit oriented
• SONET, ATM
• Works well
• Cost effective
• Will not change overnight
• Content-provider media in similar situation
• Radio, TV
• Vast installed base of equipment
• Service providers are not going to discard what
  they own and know how to operate J. McQuillan
• And what is making them money

71
The present situation (continued)

• But the Internet continues to grow, and needs a
  new infrastructure
• Which may be able to support or deliver content
  now delivered separately
• Most of todays public network revenues and
  profits do not come from the Internet

72
Convergence at transport layer

73
Convergence at switching layer

74
Convergence at application (content) layer

75
Example of convergence at application layer
Sources Beville, 1988 Time, 20 October 1997
76
Convergence at IT/Telecom layer

77
History of convergence efforts

• 1960s conversion of telephone system to digital
  technology time-division multiplexing
• 1975 SBS The One
• 1982 Integrated voice/data PBX
• 1988 ISDN
• 1991 BISDN
• 1998 Sprints ION
• 2003 VoIP, IP PBX

78
History of efforts to implement convergence

• Transport
• 1960s Conversion of telephone system to digital
  technology time-division multiplexing
• 1980s SONET
• 2000s
• All optical networks (?)
• Status Successful

• Switching
• 1975 SBS The One
• 1982 Integrated voice/data PBX
• 1988 ISDN
• 1991 BISDN
• 1998 Sprints ION
• 2003 VoIP, IP PBX (?)
• Status Has not lived up to promises

Application 1990s Delivery of audio,
graphics, text via Internet 2000s Home
entertainment/video (?) Status Evolving full
impact not yet clear
IT/Telecom 1990s Network (grid) computing,
e.g. SETI_at_Home ASPs 2000s Commercial grid
computing (?) Status ASPs successful other
areas not yet clear
79
Drivers for convergence
80
Drivers behind convergence (benefits)

• Economies of scale in service delivery
• Simplified management
• Variety of end user interface equipment minimized
• Simplified provisioning (cabling, etc.) at user
  site
• Ability to define and deliver new service types
  much greater and faster

81
Impediments to convergence

• Business/organization changes required may be
  great
• Human factors time required for people to adjust
  to new ways of doing things
• Can be measured in generations
• Deployment time (time to depreciate and replace
  existing equipment and infrastructure)
• Regulatory and legal issues

82
Impediments to convergence

83
Forces acting on convergence

84
Major changes likely to result from convergence

• Improved, faster access
• Specific defined services such as T1, Switched
  Voice, FR, likely to fade as service definition
  moves to edge of network
• New services quickly designed and rolled out
• Business models in many industries will have to
  be changed
• Current content delivery paradigms
• Infrastructure to support content delivery
• More outsourcing as network increasingly viewed
  as resource

85
Broadband penetration in home market
86
Expected rollout sequence for convergence

87
Non-technological factors expected to slow
deployment of converged solutions

• Regulatory and legal issues the existing
  telephone system is part of a web of subsidies
  and social programs which cannot be turned off
  overnight
• Human factors people have to change their habits
  and way of thinkingsomething which can require
  generations
• Economics of replacing existing infrastructure
  nothing will be replaced just for the sake of
  technological elegance there has to be a
  business case
• Proof of long-term reliability and robustness
  hard-headed CIOs and CEOs will demand to see the
  backup data, case studies, and histories
• Demonstrated cost effectiveness When all the
  costs are added up, will it still be cheaper to
  go with a single infrastructure

88
Non-technological factors expected to slow
deployment of converged solutions (continued)

• Development of new billing and support systems
  Anyone in the business knows that this is an
  extremely formidable problem. Existing systems
  work well, but developing new ones that work as
  well will be very expensive and time-consuming
• Need to interoperate with existing (legacy)
  infrastructure New technology has to be able to
  integrate and interoperate with existing
  infrastructure, both within and without a
  companys boundaries.

89
Recommended steps

• Create a strategic vision for the future of the
  networks and services that will help
  organizations achieve their business objectives.
  
• Develop a business case to rationalize the
  investment and returns
• Include soft benefits
• Do not overlook hidden costs and risks
• Identify the correct stage in the technology life
  cycle when the benefits of implementing
  convergence are maximized and the risks are
  minimized
• Not likely to be in early stages
• Develop a gradual migration strategy to integrate
  existing legacy and converged infrastructures
• Back-up strategy should be in place to mitigate
  risks that go with an evolving technology

90
Recommended steps (continued)

• Carefully examine the experiences of other
  similar organizations which have implemented the
  technology
• Determine how successful the technology is
• Whether it has met performance and business
  expectations.
• Determine external factors which may impact
  project success
• Human factors
• Cultural clashes
• Regulatory issues
• Legal matters
• Evolving standards
• Stability of companies manufacturing products
• Availability of second sources

91
Lessons learned from past

• Festina lente (Augustus)
• Technology tends to be hyped
• Non-technological factors govern if and when
  acceptance will happen
• Drop-in replacements easier and faster than those
  requiring societal reorganization

92
Where the opportunities are for optical networking

• Access (last mile)
• Necessary if really broad broadband ever to
  reach homes (10 Mbps or greater)
• Metro
• High speed network connectivity to businesses
• Debate between SONET and Gigabit/10 Gigabit
  Ethernet
• Long-haul
• Buy fibers, equipment, and networks for pennies
  on the dollar
• Global Crossing
• Enterprise

93
Optical access alternatives
Source R. Lin/Terawave Networks/NGN2000
94
Passive Optical Networking
Source R. Lin/Terawave Networks/NGN2000
95
Verizon FIOS
96
Verizon FIOS (continued)

• Pricing for FiOS
• 5 Mbps/2 Mbps for 34.95 a month as part
  of a calling package, or 39.95 a month
  stand-alone
• 15 Mbps/2 Mbps for 44.95 a month as part
  of a calling package, or 49.95 a month
  stand-alone
• 30 Mbps/5 Mbps at 199.95
• All FiOS Internet Service packages include
• Free installation by Verizon professionals
• Ultra high-speed Internet access
• Free networking router
• MSN Premium Internet Software a 99
  annual value
• 24/7 live technical support
• Up to nine verizon.net e-mail accounts
• Access to newsgroups
• Personal Web space of 10 MB
• TV also available

97
Home access speed trends
98
Long haul

• Country (and world) already well covered with
  long-haul networks
• Much dark fiber
• Laying more fiber now difficult to justify
• Better to buy up assets of failed companies

99
Cost competition is brutal in long-haul
100
Metro

• Undoubtedly some opportunities exist
• Many companies already in the field
• Ethernet solution typically used

101
Enterprise

• Networks are strategic business factors
• Run mission-critical applications
• Data must be shared with many locations
• Network downtime unacceptable
• Disaster recovery of prime importance
• Can require transfer of terabytes of data
• Often look to outsource