Journey to the Center of the Internet John Kristoff jtk@depaul.edu - - PowerPoint PPT Presentation

ISOC−Chicago 2001 John Kristoff − DePaul University 1

Journey to the Center of the Internet

John Kristoff jtk@depaul.edu +1 312 362−5878 DePaul University Chicago, IL 60604

ISOC−Chicago 2001 John Kristoff − DePaul University 2

Internet as a layered architecture

• Application layer

Web, email

• Transport layer

Reliability, flow control

• Internet layer

Routing, global addressing

• Link layer

Ethernet, PPP

• Physical layer

Wires, radio, optical fiber

ISOC−Chicago 2001 John Kristoff − DePaul University 3

Meet Ms. Dana Paquette

• She has a high−speed

Internet connection

• She’s browsing the

web

• She just clicked on a

web link

• Let’s watch...

ISOC−Chicago 2001 John Kristoff − DePaul University 4

Take me to www.isoc.org

• Web site clicked is www.isoc.org
• IP doesn’t understand names
• We must convert this to an IP address
• TCP/IP software to DNS server:

"What is the IP address of www.isoc.org?"

• DNS server replies:

"www.isoc.org = 206.131.249.182"

ISOC−Chicago 2001 John Kristoff − DePaul University 5

Protocol stack, connect()!

• Create destination TCP/IP packet using:

Destination host = 206.131.249.182

Destination application = http (port 80)

• Fill in source host information

Source IP address

Source application number

Other info (we’ll return to specifics later)

• Send connection request

ISOC−Chicago 2001 John Kristoff − DePaul University 6

TCP/IP sends packet to adapter

ISOC−Chicago 2001 John Kristoff − DePaul University 7

Ethernet card encapsulates data

Could be wireless, FDDI, cable modem, etc.

TCP/IP packet goes into payload

Ethernet dest. address = gateway router

ISOC−Chicago 2001 John Kristoff − DePaul University 8

Out the card, onto the wire...

ISOC−Chicago 2001 John Kristoff − DePaul University 9

Bit by bit...

ISOC−Chicago 2001 John Kristoff − DePaul University 10

Into the walls and ceilings...

ISOC−Chicago 2001 John Kristoff − DePaul University 11

Through the patch panel...

ISOC−Chicago 2001 John Kristoff − DePaul University 12

Onto the Ethernet switch...

ISOC−Chicago 2001 John Kristoff − DePaul University 13

To the campus router...

• Peels off layer 2 info
• Router performs

lookup for IP dest.

• Forwards towards

destination network

• Decrements time to

live field

• Re−computes IP

checksum

ISOC−Chicago 2001 John Kristoff − DePaul University 14

And out the Internet router...

ISOC−Chicago 2001 John Kristoff − DePaul University 15

To the Internet towards ISOC.org!

ISOC−Chicago 2001 John Kristoff − DePaul University 16

IP ties everything together

• IP carries data end−to−end across links
• Routers examine IP layer information

They forward towards the destination

Similar to the sorting process of postal service

• Identifies both a source and destination
• Unreliable − no guaranteed delivery!
• Primary role of IP: to move packets around

ISOC−Chicago 2001 John Kristoff − DePaul University 17

The IP datagram

ISOC−Chicago 2001 John Kristoff − DePaul University 18

The case for reliability

• Sometimes the network is offered more

packets than it can handle

Can’t queue forever

Might prefer to drop packets rather than delay them

Sender can easily re−send packets

• Need a protocol to ensure reliability

The case for TCP!

Note: reliability is placed in the hands of end−points

We’ll come back to this in a minute

ISOC−Chicago 2001 John Kristoff − DePaul University 19

Congestion control and avoidance

• TCP increases transmission rate over time
• If TCP detects a packet loss it slows down
• Competing TCPs lead to fairness over time

ISOC−Chicago 2001 John Kristoff − DePaul University 20

The TCP segment

ISOC−Chicago 2001 John Kristoff − DePaul University 21

The end−to−end picture

ISOC−Chicago 2001 John Kristoff − DePaul University 22

Dana to ISOC.org TCP/IP packet

ISOC−Chicago 2001 John Kristoff − DePaul University 23

End−to−end principle

• Guiding principle of the Internet architecture
• Considers where to put intelligence
• Minimize functions and features within the

communcations system

• Need end−to−end functions anyway
• Argues against fate−sharing and network

statefulness

ISOC−Chicago 2001 John Kristoff − DePaul University 24

Is the Internet broken?

• E2E is being violated as standard practice

Network address translation (NAT)

Firewalls

Various middleboxes

• New applications are difficult to deploy
• IPv6 could shift move back towards E2E
• Architecture has probably changed forever
• ...won’t come all the way back

ISOC−Chicago 2001 John Kristoff − DePaul University 25

Anything else wrong with the ’net?

• Security, security and security

There will continue to be major issues here

Internet is based on trust relationships

Host security is hard, net security doesn’t work

• Routing table growth

Not a critical problem, but causing some concern

Increase in multi−homing casing table bloat

ISOC−Chicago 2001 John Kristoff − DePaul University 26

What’s new and exciting?

(or "the I finished too early slide")

• Wireless
• Interactive applications

Voice and games

• IPv6
• DNS
• High−speed technologies and testbeds

ISOC−Chicago 2001 John Kristoff − DePaul University 27

References

• http://www.reed.com/Papers/EndtoEnd.html
• http://www.ietf.org

RFC 2775 Internet Transparency

RFC 1958 Architectural Principles of the Internet

• http://www.nanog.org
• http://networks.depaul.edu
• http://condor.depaul.edu/~jkristof/