Transcript ICT Technology – Some Thoughts and Issues

ICT Technology –
Issues and Opportunities
Prof. Rahul Tongia
School of Computer Science
CMU
17-899 Fall 2003
Rahul Tongia, CMU
1
Topics

Trends in Technology


Internet and Telecommunications



Time to update the adage “Cheaper, Faster,
Better – pick any 2”?
Primer
How it works (or doesn’t)
Wireless


802.11 Introduction only
Spectrum and other issues
Rahul Tongia, CMU
2
ICT – To Black Box or Not?


We can cannot cover everything in this one class
(even semester!). . .
. . .But the much of the technological issues are not
that hard – despite some people wanting to pretend
they are.

With a little effort, the important details can be extracted
Rahul Tongia, CMU
3
Requirements for Successful Service
Will it inter-operate?
Can it be built?
Technology
Standards
Regulation
Is it allowed?
Rahul Tongia, CMU
Market
Will it sell?
4
Industry & Society: Penetration Rates
Users (Millions)
Years to reach 50M users:
120
90
Radio
TV
= 38
= 13
Cable
= 10
Internet
=5
60
30
Radio
Cable
TV
Internet
0
‘22
‘30
‘38
‘46
‘54
‘62
‘70
‘78
‘86
‘94 ‘02
Source: Morgan Stanley
Rahul Tongia, CMU
5
The Heart of the Matter: The Growth of Computers
1638400
819200
Tera PC
Doubling every 15 months
409600
204800
102400
100G PC
51200
M25600
I
P12800
S 6400
Doubling every 2 years
10G PC
3200
1600
Giga PC
800
From: Raj
Reddy- The
Global Village
400
200
100
1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020
Year
Rahul Tongia, CMU
6
Storage Performance
Rahul Tongia, CMU
7
Optical Fiber: Promise & Performance
1,000Gb/s
1995
Projection
100 Gb/s
1987 Projection
1983 Projection
10 Gb/s
1980 Projection
1 Gb/s
1978 Projection
1980
1990
2000
Bell Labs
Gilder’s Law – Optical speeds doubling in ~ 9 months
Rahul Tongia, CMU
8
Software Challenges in Intelligent Data Processing
User Decision Support Demand
vs. Processor speed
Database demand:
2X / 9-12 months
100
“Greg’s Law”
10
1
1996
“Moore’s Law”
1997
1998
1999
Database-Proc.
Performance Gap:
CPU speed
2X / 18 months
2000
38
D. Patterson & Kimberly Keeton
UCB
Comparative Statistics
Rahul Tongia, CMU
CAIDA (2002)
10
What Makes the Internet tick?

The Internet runs on 3 things:

Boundaries



Limits of Responsibilities
Inside the core, is like a black box (“The Cloud”)
Standards (protocols) for data-centric design

Expectations of how things should work together


Robustness Principle





Layering
"Be liberal in what you accept, and conservative in what
you send.“ – Jon Postel
Resiliency – distributed architecture
Limits Monopolies
NO ONE OWNS THE INTERNET
Trust


Addressing schemes and registration
End-to-end design
Rahul Tongia, CMU
11
What is the Internet?





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
Rahul Tongia, CMU
a.k.a. Backbone Providers
Tier 1
Tier 2
Users
• There are often more layers
• There can be interconnections other than
12
at a backbone
Structures of the Industry


Government Dept.
Government company (PTT)



PTT: Abbreviation for postal, telegraph, and telephone
(organization). In countries having nationalized services,
the organization, usually a governmental department,
which acts as its nation's common carrier.
Regulated Monopoly
Competition



IXC – Inter Exchange Carriers
ILECs – Incumbent Local Exchange Carriers (Baby Bells)
CLECs – Competitive Local Exchange Carriers


Overbuilders
Unbundled Network Elements (Open Access)
Rahul Tongia, CMU
13
“Call Completion” / Transaction Charges

Mail – postage stamp mechanism
Telephony – cost sharing mechanisms (vary)
Internet?

What are the costs?






Calling – sharp falls over time
Mailing – increasing over time
Faxing – not going away anytime soon
Email

Is it really free?



Rahul Tongia, CMU
Access
Upstream TCO (ignoring SPAM, for now!)
Time
14
Peering – Internet “Call Completion”

Where backbones come together


Major design issue (relates to cross-connection)
Public Peering – fallout of the public history of
the Internet

Network Access Points (NAPs)


Started with 4, but now there are more
Usually done by equals


Private Peering


Give as much traffic as receive
Commercial (private)
International peering is more limited (links
are much more expensive)
Rahul Tongia, CMU
15
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
Routers pass packets along towards the
destination hop-by-hop
Rahul Tongia, CMU
16
Internet – Good for what it was made for

Best-effort data network



Scalable
Resilient
New trend – Everything over IP (XoIP)

Voice – Circuit switched

Less than half the traffic


But, is most of the revenue for carriers




Growth of ~25% vs ~100% (?) for data
Suppliers’ “killer app”
For users, email and WWW are the killer apps (legal,
anyways)
Internet Telephony is not the same as VoIP
Latency example

Berkeley – CMU IP-based lectures!
Rahul Tongia, CMU
17
Applications vs. Networking Parameters
Rahul Tongia, CMU
18
Internet is built on trust:

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
Rahul Tongia, CMU
19
Standards and Regulation

Many bodies, sometimes with overlap




IETF (within IAB) 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
Rahul Tongia, CMU
20
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


NSF stopped paying NSI, allowed NSI to
charge for .com, .net, .org


Network Solutions, Inc. (NSI) [under InterNIC]
$70 for two years
NSI becomes enormously profitable
Rahul Tongia, CMU
* Based on information from Jon Peha and Gary Kessler
21
Domain Names (cont.)

NSF responsibilities passed to Commerce Dept.


NSF establishes ICANN (Internet Corporation for
Assigned Names and Numbers) in 1998


Has many critics
Registration became competitive by 1999



The US government controlled key element of the Internet
(!) so
Registry: manage database, NSI monopoly
Registrar: consumer interface, competition
IP address space (numeric) is still from
regional authorities
Rahul Tongia, CMU
22
Spectrum

Frequency affects




Capacity 
Bandwidth
Range
Interference
and Line of
Sight
Requirements
Protocols and
Technology
ISM Bands are
kept free for
Industrial,
Scientific, and
Medical
Applications,
e.g., 2.4 GHz
Rahul Tongia, CMU
23
Special Properties of Spectrum

Heavily controlled






Military uses
Licensed use
Source of licensing fees
Is a public good; everywhere yet not
limitless
Many forms are appropriate for point to
multipoint (including broadcast)
Encoding is key – bits per hertz
Rahul Tongia, CMU
24
Spectrum Issues

802.11 Alphabet Soup

a, b, g, i, etc. – Differ in





Is licensed spectrum better (cleaner, scalable,
etc.)?


Data Rates
Bands
Compatibility
Distance
3G licenses have gone for thousands of dollars per
potential subscriber
Cognitive Radios might be the future
Rahul Tongia, CMU
25
Hypothetical WiFi Kiosk

Access Points are now about $100 (only!)


What else does it take?
What range does it cover?

Number of Users




FCC vs. ETSI regulations on emissions
Uplinking


Band overlaps and congestion
IP address space
“Now What” Syndrome – need user h/w, s/w, etc.
Business Plan ?

Capex is less than half of “broadband” costs
Rahul Tongia, CMU
26
ICT Issues

Policy




Universal Service / Digital Divide
Globalization


Convergence
Open Access
“Winner Takes All”
Internet


Is it special (Information Service vs.
Telecom Service)?
Jurisdiction

Taxation
Rahul Tongia, CMU
27
Issues in the Internet

Scalability


Internet is growing* at 100-300%
Running out of IP addresses – esp. LDCs

Long term solution: IPv6



Protocols and equipment are straining
Security





128 bit addresses (millions per square meter)
Distributed Denial of Service – example of an attack
Viruses
Spam
Privacy
Quality of Service

Voice
Rahul Tongia, CMU
28