Internet Overview

1
Internet Overview Accessing Information
Digital Divide

• March 6, 2008

2
Design and History of the Internet

• Layperson misconceptions
• WWW Internet Email online broadband
• Some questions to think about
• Who owns the Internet?
• Who controls the Internet?
• Is the current system OK?
• Security
• Scalability
• Usability

3
Structures of the Telecom Industry

• Government Dept.
• Government company (PTT)
• Regulated Monopoly
• Competition
• Splits within sectors
• IXC InterExchange Carrier (Long Distance)
• ILECs Incumbent Local Exchange Carrier (Baby
  Bells)
• CLECs Competitive Local Exchange Carrier

4
Government Departments

• Losing ground
• Privatization big push
• Type 1
• Public Assets privatized and then regulated
• Type 2
• Government carrier becomes one of many players

5
PTT

• PTT Abbreviation for postal, telegraph, and
  telephone (organization). In countries having
  nationalized telephone and telegraph services,
  the organization, usually a governmental
  department, which acts as its nation's common
  carrier.

6
Call/Transaction Completion Charges

• Mail
• Flat Rate
• Telephony
• Usage based or flat rate
• Internet?
• Depends on what user (residential, commercial,
  bulk, etc.)

7
What is the Internet?
a.k.a. Backbone Providers

• The global (public) network built from hundreds
  and thousands of internetworking independent
  networks.
• No single entity runs the Internet
• Operates on standards
• Built on a modified hierarchical structure
• Packet Switching

Tier 1
Tier 2
Users

• There are often more layers
• There can be interconnections other than at a
  backbone

8
What makes the Internet the Internet?

• Open architecture
• Standards and protocols allow applications and
  communications without caring of the underlying
  infrastructure or system
• The Cloud
• Anyone can access anything (is public)
• Resiliency (mesh design)
• End to end system

9
How big is the Internet?

• Many metrics
• Number of Service Providers
• Number of Hosts
• Number of Subscribers
• Size of Interconnections
• (see outside sources such as CAIDA, Hobbes
  Internet Timeline, etc.)

10
Brief History of Internet Evolution

• 1969 ARPANET 50 kbps UCLA, UCSB, SRI,
  and Utah
• 1970 56 kbps transcontinental adding BBN,
  MIT, RAND
• 1972 50 kbps 23 hosts
• 1973 75 of traffic on ARPANET is email
• 1981 CSNET (in parallel) 56 kbps 213 hosts
• 1983 TCP/IP mandatory, DNS created 562 hosts
• 1985 NSFNET initiated 1.544 Mbps 1961 hosts
• 1987 UUNET created for commercial access
• 1990 ARPANET disbanded in favor of
  NSFNET 313,000 hosts
• 1992 NSFNET 45 Mbps upgrade complete 1,136,000
  hosts
• ( a few pvt. Backbones)

11
Brief History of Internet Evolution (cont.)

• 1994 NSFNET 145 Mbps ATM 3,864,000 hosts
• ( a few pvt. Backbones of 56 kbps, 1.5 Mbps,
  and 45 Mbps)
• 1995 NSFNET privatized to 4 players 6,642,000
  hosts
• 1996 MCI 622 Mbps
• 1996 - Now upgrading to 2.5 and 10 Gbps IP
  links
• This history has helped shape US Internet
  architecture in terms of competition and layout
  (peering)

12
Peering

• Where backbones come together
• Major design issue (relates to cross-connection)
• Public Peering
• Network Access Points (NAPs)
• Started with 4, but now there are more
• Usually done by equals
• Give as much traffic as receive
• Private Peering
• Commercial (private)
• International peering is more limited (links are
  much more expensive)

13
Open Systems Interconnection (OSI) Model
examples
Interface MESSAGES User Interacts with these
FTP, Ping, HTTP, etc.
Translation and encryption MESSAGES
Remote Procedural Calls (RPCs), Error Checking
MESSAGES
Reliability, Error-checking SEGMENTS end-to-end
validity
TCP
Software Address, Routers DATAGRAMS establishes
routes (extends nodes)
IP
Hardware Address, Bridges, Intelligent hubs,
NICs, Error Checking FRAMES node-to-node
validity
Ethernet, ATM
Pins, Wires, Repeaters, RS-232, Volts, etc
BITS Deals with the medium
SONET/SDH
14
Ethernet

• A standard for networking at Layer 2
• Based on physical hardware address (12 Hex
  numbers)
• First started within the LAN
• Started of as a shared bus (from the Aloha Packet
  Radio network Bob Metcalf)
• New versions are full-duplex, switched
• Amenable for optical, longer reach
• Graceful evolution (backwards compatible) between
  10/100/1000 Mbps
• Ethernet Frames are between 64 and 1518 bytes in
  size
• IEEE is the standards body (802.xx working groups)

15
Ethernet Operation (traditional)

• Carrier Sense Multiple Access/Collision Detect
  (CSMA/CD)
• All machines wait to see if medium is free
• If so, they transmit
• Sometime, packets can collide
• In that case, the transmitters wait a random
  period of time, and re-transmit
• If yet another collision, will wait longer period
  of time (exponential back-off)
• Limitations
• Effective bandwidth was modest
• Distances were limited
• Non-duplex

16
TCP/IP

• Suite of protocols for networking
• Based on logical address for devices
• Most popular standard worldwide built into most
  OS
• Like most other packet switching, is
• Connectionless
• Statistical (non-deterministic)
• No inherent Quality of Service (QoS)
• Most of IP routing is unicast
• Packets carry lots of information
• Source Address, Destination Address, etc.
• Special instructions such as priority
• Port number (meaning application ID)
• E.g., Port 80 - http

17
IP Addresses

• Each device connected needs a unique IP address
• Exception is private IP addresses used within
  non-global networks
• Home gateways can use this
• Gateway router translates between public and
  private IP addresses
• 32 bit addresses in current version (IPv4)
• 4 8-bit portions
• Dotted decimal is popular for convenience
• 128.2.72.44 is same as 10000000.00000010.01001000.
  00101100

18
IP Addresses (cont.)

• IP addresses have 2 portions, network and host
• Networks are uniquely controlled. e.g, 128.2.x.y.
  is CMUs network
• Earlier, IP addresses were class-based to
  differentiate
• Newer system is classless can arbitrarily
  demarcate network and host
• A.B.C.D/24 implies first 24 bits are for network
  portion
• More efficient
• Subnet Mask is used to identify network portion
• Most people dont own their own network they
  take a portion from their service provider

19
Network boundaries

• LANs used to predominate
• Old rule of thumb 80 traffic inside 20 outside
• Often were Layer 2 networks
• Intranet
• Can make an outside, non-global network
• Extranet
• Often using private (leased lines)
• Outside world
• Layer 3 connections (IP)
• Many types of interconnections, e.g., varying by
• Speed
• Dial-up
• Dedicated connection Just a pipe to the cloud
• Protocol
• IP, IPX, Appletalk, etc.

20
Routers

• Forward packets based on destination address
• They know the route to every network
• Once the packet gets to the network gateway, it
  internally finishes the routing
• Todays Internet is roughly 200,000 routes in
  size (advertised prefixes 2006 estimate)
• Routing is done on a hop-by-hop basis
• A routing table is built up in each router
• Incoming packets destination address is looked
  up
• A match is made, and the packet is forwarded to
  the appropriate port which gets it one step
  closer to the destination

Incoming packet for 128.2.x.y
128.4.x.y
Router
A
C
Routing table knows which port (interface) is
most closely connected to a particular network(s)
D
B
128.2.x.y
128.3.x.y
21
IP Routing

• Core Routing
• Internet-sized routing tables
• Optical interfaces
• Edge Routing
• Traditional edge players (aggregators)
• Metropolitan Area Network/GigE edge players
• Wide Area Networking is different from LAN, even
  though many protocols are the same
• Access (Customer Edge)
• Often the bottleneck
• Earlier, relied on the ILEC (e.g., Verizon)
• Now, new carriers want to bypass the ILECs
• Often use new technologies and standards

22
Communications Components

• Transport
• Now, typically optical, except the last mile
• Termination
• Different devices (typically) for different
  layers
• Phones, Video-conf. phones, routers, modems, etc.
• Switching
• Cross Connects / Add-drop Multiplexers (ADMs)
• Class 4/5 switches
• IP switches (Routers)

23
Network Intelligence

• Quality-of-Service (QoS)
• Todays Internet is best-effort
• Need to differentiate different packets
• Issues of identification, authentication, and
  billing
• Critics content some schemes amount to violation
  of Net Neutrality
• Moving Intelligence to the Edge
• Filtering, monitoring, and differentiating
• Lets the core be super-fast
• Security
• Todays internet is inherently insecure
• Higher layers are used for security
• E.g., SSL in browswers
• New designs are being worked on for more security

24
(No Transcript)
25
What do People Access in the Last Mile?

• Voice
• Video
• Broadcast
• Switched
• Even On Demand
• Broadband Internet Access
• These are the TRIPLE PLAY

26
What do People Access?

• (Mid 2000s)
• Predictions were p2p would only grow
• Something changed
• VIDEO (e.g., YouTube)

Source CacheLogic
27
IPTV Bit Rates
Source http//www.dslprime.com/pix/cbrrates.jpg
28
Broadband AccessThe Last Mile

• Different technologies are available
• Cable
• DSL
• Fiber
• Wireless
• Fixed
• Mobile
• Satellite
• Powerline
• They differ in
• Reach
• Speeds
• Costs
• Regulation (?)

29
Cable Hybrid Fiber Coax (HFC)
Active
Node
Home
Headend
Feeder (Fiber)
Drop Loop
Source Marvin Sirbu
30
Advanced Hybrid Fiber Coax
Coaxial Termination Unit
Active
Node
Home
Headend
HDT
Feeder (Fiber)
PSTN
Drop Loop
Source Marvin Sirbu
31
CABLE MODEMS
CMTS
O/E
O/E
Video
set top
fiber node
Head End
T
Spectral Use
10 BaseT
IAP
U P
T V
T V
T V
T V
D O W N
Cable Modem
900M
750M
0
50M
Internet Backbone
Frequency
optoelectronics
O/E
Source Stagg Newman
32
DSL from Central Office
Subscriber Premises
Central Office
ADSL Modem
PC
Voice Switch
DSLAM
Data carried above 4KHz voice frequencies
Splitter
Data Switch
Telephone
This simplification ignores the use of remote
terminals and digital loop carrier (DLC)
Source Marvin Sirbu
33
Fiber to the Neighborhood
Central
Distribution
Drop
Office
Plant
Plant
ADSL
Manhole
Fiber Optic
Central
Feeder Plant
Inter-Office
Office
Trunking
Local Access Network

• Can go all the way to the home (FTTH)
• Fiber can easily provide Gigabit speeds

Source Marvin Sirbu
34
VDSL vs ADSL
Source http//www.comsoc.org/comsig/Slides/Oct20
03_DSL_BernardDebbasch.pdf, Oct 2003
35
Distance vs Bit Rate and Video Delivery
Source http//www.aware.com/products/dsl/bonded.
htm
36
Challenges with Wireless

• What prevents us from more wireless broadband?
• Spectrum
• Reach
• Related to power levels
• Line of Sight
• Costs
• Evolving standards and technologies
• WiFi
• Mesh, MIMO, etc.
• WiMax
• Fixed and Mobile
• 3G, 4G, etc.

37
Fixed Wireless Access Inherently Shared

• Base station
• Point to Multipoint
• Receivers
• Rooftop
• Indoors
• Mobile/Portable
• Shared bandwidth depends on technology
• 25-40 Mbps downstream (might be)
• 15-25 Mbps upstream
• Spectrum matters
• Unlicensed (UNI 5 GHz)
• Licensed (e.g., MMDS - 2.5 GHz)

38
MMDS Fixed Wireless Architecture Base Station
and CPE
Transceiver/Antenna
Adapter
Other MMDS channels
Tower and Antenna (Base Station Outdoor Unit)
Wireless Modem Unit
Wireless Modem Termination System
Transceiver/Antenna
Router/ ATM switch
Base Station Indoor Unit
Wireless Modem Unit
Fiber Backhaul To Distribution Hub
Ethernet LAN

• Sprint and MCI have purchased extensive MMDS
  licenses and will roll out in 40-60 markets over
  the next year.

VoIP Adapter
Small Business
Source Marvin Sirbu
39
Customer Fixed Wireless Units

• Typically, requires clearLine of Sight (LOS)
• Except in small radius
• This requires costly site visit to install
  antenna, run wiring to computer
• Newer alternatives emerging (non-LOS)

Source Sprint (Hybrid Networks) (antenna/transce
iver only)
40
Base Station Equipment

• A single tower can cover up to 20 mile radius
• Depends on terrain
• As subscribers increase, may need additional base
  stations/cells for frequency reuse

Source Sprint (Hybrid Networks- Phoenix)
41
Wireless ISPs

• There are several thousand Wireless ISPs (WISPs)
  in the U.S.
• Easy because of light touch regulation
• Spectrum
• Antennae
• Majority of WISPs use souped up wireless LAN
  technology
• Normal WLAN coverage few hundred feet
• With directional antennas, coverage can reach
  several miles

42
Wireless Mesh Networks

• Popular for many city networks
• Philadelphia, San Francisco, etc.
• Major advantage
• Issues of backhaul
• Challenge
• Shared throughput
• Business model questions
• Free vs. subsidized vs. at cost
• Q Can one share connectivity?
• Open Access Points (mesh or non-mesh)?

43
Antennas for Long Range WLANs
Source Cisco
44
Broadband Policy Issues

• Unanswered questions
• Is there a natural monopoly in broadband?
• Very low marginal costs in telecom
• How can one support competition over broadband
  infrastructure?
• Who should build broadband networks?
• Public/Private
• Market/Regulated
• How do we define and pay for Universal Service?
• Thinking of layers or boundaries becomes
  important
• Wholesale vs. retail
• Physical vs. logical
• Content vs. carriage

45
What is changing?

• Applications
• Protocols
• Peer2Peer why is it popular
• Size of files
• New Killer apps
• Where we access information
• On the move

46
Sometimes, its all About the
47
Components of Connectivity
Hardware / Installation
Marketing / Advertising
OM
Uplinking (transit fees)
Technical
CRM

• Vary by location
• Oversubscription ratios are an ISP choice
• Speeds offered determine what applications can
  be run

• Varies by technology

• Also depends on competition

• One time costs
• Depends on competition

• One-time capital costs are amortized over time
• Cost depends on
• Interest rates
• Churn
• Re-usability of components

48
What does it Cost to use up Bandwidth?
/Mbpstransit
StatisticalMultiplexing(oversubscription)
Mbps uplinked
Number of userssharing a link
/month costper user to ISPfor uplinking
Rated Bandwidth
49
Different Bits are Different
p picodollars 10-12

• Voice
• Fixed
• 23 /month, 1 month/1923 min. ? 3,100 p/bit
• LD
• 0.10/minute ? 26,000 p/bit
• Incl. International charges (FCC numbers)
• Web (broadband user)
• 35 /month, 2 hours per day usage, 30 kbps
  average usage ? 5,400 p/bit
• TV (cable/satellite, excl. PPV)
• 225 /year/person, 2.58 persons/household, 850
  hours/year watched ? 36 p/bit
• A good fraction of their revenues comes from
  advertising
• BUT, we dont know what demand will look from 5
  years from now, or, say, under 100 Mbps conditions

2002 or 2003 US Statistical Abstract Average
Numbers except in Italics
50
Digital Divide
51
The 4C Framework

• Information and Communications Technology (ICT)
  can be thought of as the 4Cs
• Computers
• Devices
• Connectivity
• Analog/digital packet/circuit
• Content
• Centralized/decentralized
• (human) Capacity
• Literacy, language, etc.

52
US Broadband Penetration

• Why is this misleading?

53
Global Broadband

• Why could such information be misleading?

54
Truer Picture of Global Broadband

• Issues of speeds or price are not shown

55
Digital Divide
Source CAIDA
56
What is the Digital Divide?

• Digital Divides are not just the result of
  economic differences in access to technologies
  (Haves vs. Have-Nots), but also in cultural
  capacity and political will to apply these
  technologies for development impact (Dos vs. Do-
  Nots).
• Markle Foundation Report (2003)

57
What is the Digital Divide?

• The divide is a manifestation of underlying
  divides, a symptom rather than a cause
• Economic, social, gender, age, geographic, etc.
  divides
• It is a moving target
• Dial-up, broadband, real broadband, etc.
• Information fuels the present (Knowledge)
  Revolution
• Enables the Drivers of Growth
• Access ? Information ? Knowledge ? Opportunity

58
There are other Metrics and Divides
US may rank 19th in broadband (2005), but
Newer data indicate the US is now 53rd!
59
4 Dimensions of the Digital Divide

• Awareness
• What is it, and what can one do with it?
• Availability
• Is it offered to me?
• Accessibility
• Can I realistically use it (including issues of
  literacy and language)?
• Affordability
• Globally, ICT is 6.6 of GDP (telecom, hardware,
  and software)
• What percentage of income does access cost
  worldwide?

60
Improvements are needed in all Dimensions of ICT

• Computers
• Life cycle analyses
• Interface
• Connectivity
• Broadband?
• Content
• Locally relevant information
• (human) Capacity
• Literacy
• e-Literacy

61
Why is Connectivity so Expensive in Developing
Countries?

• Issues of scale few users
• International Gateway bottlenecks
• Licensing fees and duties
• Monopoly carrier (de-facto, often)
• Poor design
• And many more reasons

62
Mobile Phones

• Dominant connectivity in much of the world
• 10 penetration in Africa!
• Largest market in the world today is?
• BUT, the Avg. Rev. Per User (ARPU) can be high
  (expensive)
• Africa (2004) 28
• India was only 11 (and under 8 today)
• Do mobiles have data capabilities?

63
100 Laptop Pros and Cons

• Pros
• Creates awareness
• Might have some innovation
• In some cases, may fulfill a latent need
• Cons
• Top-down
• Robustness unknown
• Energy
• Connectivity
• Wont share easily
• Buy-in is expensive
• Content?
• Role of teachers

64
Idea FiberAfrica Concept

• A revolutionary design to provide the majority of
  the population nearby access to broadband for a
  one-time capital expenditure of 1/capita
• Can be cheaper by harnessing any existing
  infrastructure
• Includes optical fiber of virtually unlimited
  capacity between major population centers, and
  broadband wireless hubs for wide-spread access
  over large areas
• Excludes PCs and end-user equipment
• Revolutionary business model could allow
  virtually free access to schools, hospitals and
  rural community centers

65
FiberAfrica Backbone Network

• Almost 70,000 km core backbone (shown)
• DWDM Technology for scalability and
  cost-effectiveness
• 35,000 km fiber spurs (not shown)
• Routing chosen to provide maximum coverage
• Can leverage existing fibers and rights of way
  (along highways

66
FiberAfrica Design
Detailed design undertaken, for all capital and
operating expenses
67
Business Model(s)

• Many options available, but requirements include
• Operational costs must be covered
• Our calculations show it can be done, affordably
• Capital costs can be grant-based (only 1B)
• There must be end-user and community empowerment
• Public core, competitive edge
• Wont create a new government (or other)
  bureaucracy
• Consortium or partnership models have worked,
  e.g., IntelSat
• Allows role for AfricaUnion/NEPAD as appropriate
• Maintains individual governmental sovereignty

68
Why This Model?

• Appropriate scale into the rural areas
• Optical fibers make it future-proof
• One time cost leads to fiber infrastructure that
  can last decades
• Capital costs of fiber much lower than
  conventional wisdom
• Few thousand /km maximum
• Increases access and domestic usage not
  addressed merely by having an international fiber
  link (e.g., EASSY proposal)
• Closed Club arrangements of such fiber systems
  make them unaffordable
• Business model is sustainable
• Public-Private partnership
• Synergistic with mobile providers who lack such
  capacity for broadband
• Almost no barrier to entry for casual users
  (through schools and community access points)

69
Open Access / FiberAfrica Underpinnings

• Overcoming the infrastructure disconnect
• Fiber lasts 30 years, electronics need to be
  amortized in 5-7 years
• Today, carriers often charge more (short-term
  business models)
• Higher cost models are inherently a niche
  solution
• No conflict with competition
• Focus on rural and uneconomic areas
• ISPs would also benefit
• Can justify special regulation only for the
  public good
• Could also attract grants and soft loans

70
More info on FiberAfrica

• For more information, see
• http//www.contrib.andrew.cmu.edu/tongia/FiberAfr
  ica--ending_a_digital_divide.pdf
• OR
• http//tinyurl.com/dttga

71
(No Transcript)
72
Internet is built on Principles, not Laws

• Registration (databases) are believed because
  people think they are correct
• Domain Name System
• Handles names for humans vs. binary for machines
• Root names are the last .xxx, e.g., .com, .edu,
  .org, .mil, .ca, .tv
• Just 13 root servers in the world
• Many copies made for practical purposes
• Borders define responsibilities
• Best effort (democratic)
• Robustness
• "Be liberal in what you accept, and conservative
  in what you send.
• - Jon Postel

73
Standards and Regulation

• Many bodies, sometimes with overlap
• IETF handles the engineering of the network
• W3C handles web standards such as html, xml, etc.
• IEEE handles some standards
• Requests for Comments (RFCs) are how things get
  standardized
• Draft is circulated
• Modified, debated, etc. (many versions often)
• Becomes a standard by vote.
• Companies often try and tilt emerging standards

74
Registries and Domain Names

• Numeric address space is coordinated
• Domain Names initially managed by ISI (Jon
  Postel)
• National Science Foundation (NSF) hired
  contractor to administer
• Network Solutions Inc (NSI)
• NSF stopped paying NSI, allowed NSI to charge for
  .com, .net, .org
• 70 for two years
• NSI becomes enormously profitable
• NSF responsibilities passed to Commerce Dept.
• The US government controlled key element of the
  Internet (!) so
• NSF establishes ICANN (Internet Corporation for
  Assigned Names and Numbers)

Based on information from Jon Peha
75
Domain Names (cont.)

• ICANN decisions
• Protect trademark owners
• Oppose cybersquatting
• Do not create more top level domains
• Divide NSI responsibilities
• Registry manage database, NSI monopoly
• Registrar consumer interface, competition
• NSI claims to own the .com, .net, .org database
• Do they have to give it up or share it?
• ICANN says that NSI must be accredited
• NSI refuses to sign agreement with ICANN
• NSI does not recognize ICANN's authority
• NSI protects its revenue stream
• What happened in the end?
• NSI was acquired by VeriSign, then spun off

76
Domain Names (cont.)

• ICANN critics
• NSI and friends, many academics
• ICANN is the evil face of governance in the
  Internet, which needs no governance
• ICANN is an unrepresentative, unelected group
  with unlimited power
• Rest of World (especially developing countries)
  particularly dislike the entire process (not just
  ICANN)
• Meet behind closed doors, create taxes ...
• ICANN supporters
• ICANN, many high-tech companies, trademark
  owners.
• NSI is an unregulated monopoly that must be
  stopped.
• Engineers seeking consensus, do not address
  policy.
• A neutral group of experts making necessary
  decisions.
• ICANN people are just "plumbers
• Remains a major issue Internet Governance
• What is the debate about?

77
Issues in the Internet

• Scalability
• Internet is growing at 75-300
• Running out of IP addresses
• Long term solution IPv6
• 128 bit addresses (millions per square meter)
• Protocols and equipment are straining
• Security
• Distributed Denial of Service are an example
• Viruses
• Quality of Service
• Voice
• Usability

78
Issues in the Internet (cont.)

• Privacy
• Anonymity
• Identity
• Regulation
• Universal Service Obligation
• Taxation
• Encryption (and its a technology issue)
• Digital signatures
• Digital Divide

79
Policy Issues (Discussion)

• Are Terms of Service sufficient to disallow
  Domain resolution?
• E.g., GoDaddy vs. Seclists.org dispute over
  MySpace complaint
• How do we do CALEA on the Internet?
• Can we?
• Should we?
• What about Skype?
• Is not a phone service, but a voice IM (?)

80
(No Transcript)