Sie sind auf Seite 1von 41

Internet architecture and

history

Introduction 1-1
Internet Architecture
q http://www.nap.edu/html/coming_of_age/
q http://www.ietf.org/rfc/rfc1958.txt

Introduction 1-2
Why did the Internet win?
q Packet switching over circuit switching
q End-to-end principle and “Hourglass” design
q Layering of functionality
q Distributed design, decentralized control
q Superior organizational process

Introduction 1-3
Packet switching versus
Circuit switching
q Analogy
v Zip cars vs. privately owned cars

q Zip cars
v Many users share a single car
v Large demand for cars causes users to delay
usage
v Car is more efficiently used

q Privately owned cars


v Single user
v Guaranteed access for user
v Car is not used as efficiently

Introduction 1-4
Packet vs. circuit switching
q mesh of interconnected
routers
q the fundamental
question: how is data
transferred through net?
v circuit switching:
dedicated circuit per
call: telephone net
v packet-switching: data
sent thru net in
discrete “chunks”

Introduction 1-5
Circuit Switching

End-end resources
reserved for “call”
q network resources
(e.g., bandwidth)
divided into “pieces”
v link bandwidth, switch
capacity
v pieces allocated to calls
v resource piece idle if not
used by owning call
• dedicated resources: no
sharing
q circuit-like (guaranteed)
performance
q call setup and admission
control required
Introduction 1-6
Case study: Circuit Switching

q 1890-current: Phone network


v Fixed bit rate
v Mostly voice
v Not fault-tolerant
v Components extremely reliable
v Global application-level knowledge throughout
network
v Admission control at local switching station
(dial-tone)

Introduction 1-7
Network Core: Packet Switching
each end-end data stream congestion:
divided into packets q aggregate resource
q user A, B packets share demand can exceed
network resources amount available
q each packet uses full link q packets queue, wait for
bandwidth link use
q resources used as needed q store and forward:
packets move one hop
at a time
Bandwidth division into “pieces”
Dedicated allocation
Resource reservation

Introduction 1-8
Packet Switching: Statistical Multiplexing
10 Mb/s
A Ethernet statistical multiplexing C

1.5 Mb/s
B
queue of packets
waiting for output
link

D E

Sequence of A & B packets does not have fixed pattern,


shared on demand è statistical multiplexing.

Introduction 1-9
Packet switching versus circuit switching
Packet switching allows more users to use network!
q N users over 1 Mb/s link
q each user:
v 100 kb/s when “active”
v active 10% of time
q circuit-switching:
v 10 users N users
q packet switching: 1 Mbps link
v with 35 users, probability
> 10 active less than .0004
v Allows more users to use
network Q: how did we get value 0.0004?
v “Statistical multiplexing
gain”
Introduction 1-10
Packet switching versus circuit switching
Is packet switching a “slam dunk winner?”
q Great for bursty data
v resource sharing
v simpler, no call setup
q Bad for applications with hard resource requirements
v Excessive congestion: packet delay and loss
v Need protocols for reliable data transfer, congestion control
v Applications must be written to handle congestion
Q: How to provide circuit-like behavior?
v bandwidth guarantees needed for audio/video apps
v still an unsolved problem
v Common practice: over-provision

Introduction 1-11
Problems with packet switching
Packet loss and queuing delay
packets queue in router buffers
q packet arrival rate to link exceeds output link capacity
q packets queue, wait for turn
q when packet arrives to full queue, packet is dropped (aka lost)
v lost packet may be retransmitted by previous node, by source end
system, or not retransmitted at all
packet being transmitted (delay)

B
packets queueing (delay)
free (available) buffers: arriving packets
dropped (loss) if no free buffers
Introduction 1-12
Case study: Packet Switching

q 1970/80s-current: Internet network


v Variable bit rate
v Mostly data
v Fault-tolerant
v Components not extremely reliable (versus
phone components)
v Distributed control and management

Introduction 1-13
Why did the Internet win?
q Packet switching over circuit switching
q End-to-end principle and “Hourglass” design
q Layering of functionality
q Distributed design, decentralized control
q Superior organizational process

Introduction 1-14
End-to-end principle and
Hourglass design

Introduction 1-15
End-to-end principle
q J. H. Saltzer, D. P. Reed and D. D. Clark
“End-to-end arguments in system design”,
Transactions on Computer Systems, Vol. 2,
No. 4, 1984
q http://www.acm.org/pubs/citations/journal
s/tocs/1984-2-4/p277-saltzer/

Introduction 1-16
Hourglass design
q D. Clark, “The design philosophy of the
DARPA Internet”, SIGCOMM 1988, August
16 - 18, 1988.
http://www.acm.org/pubs/citations/proceedings/comm/52324/
p106-clark/

Introduction 1-17
End-to-end principle
q Where to put the functionality?
v In the network? At the edges?

q End-to-end functions best handled by end-to-end


protocols
v Network provides basic service: data transport
v Intelligence and applications located in or close to
devices at the edge
v Violate principle as a performance enhancement
q Leads to innovation at the edges
v Phone network: dumb edge devices, intelligent network
v Internet: dumb network, intelligent edge devices

Introduction 1-18
Hourglass design
q End-to-end principle leads to “Hourglass”
design of protocols
q Only one protocol at the Internet level
v Minimal required elements at narrowest point
q IP – Internet Protocol
v http://www.rfc-editor.org/rfc/rfc791.txt
v http://www.rfc-editor.org/rfc/rfc1812.txt
v Unreliable datagram service
v Addressing and connectionless connectivity
v Fragmentation and assembly

Introduction 1-19
Hourglass design
q Simplicity allowed fast deployment of multi-
vendor, multi-provider public network
v Ease of implementation
v Limited hardware requirements (important in 1970s)
• Is it relevant now with today’s semiconductor speeds?
v Eventual economies of scale
q Designed independently of hardware
v No link-layer specific functions
v Hardware addresses decoupled from IP addresses
v IP header contains no data/physical link specific
information
v Allows IP to run over any fabric

Introduction 1-20
Hourglass design
q Waist expands at transport layer
v Network layer = host to host communication
v Transport layer = application to application communication
q Two dominant services layered above IP
q TCP – Transmission Control Protocol
v Connection-oriented service
v http://www.rfc-editor.org/rfc/rfc793.txt

q UDP – User Datagram Protocol


v Connectionless service
v http://www.rfc-editor.org/rfc/rfc768.txt

Introduction 1-21
Hourglass design
q TCP – Transmission Control Protocol
v Reliable, in-order byte-stream data transfer
• Acknowledgements and retransmissions
v Flow control
• Sender won’t overwhelm receiver
v Congestion control
• Senders won’t overwhelm network
q UDP – User Datagram Protocol
v Unreliable data transfer
v No flow control
v No congestion control

Introduction 1-22
Hourglass design
q What uses TCP?
v HTTP (Web), SMTP (E-mail transmission), IMAP, POP (E-
mail access)
q What uses (mainly) UDP?
v NTP (network time protocol), Highly interactive on-line
games (First-Person Shooters)
v Many protocols can use both
q Check out /etc/services on *nix or
C:\WIN*\system32\services
q IANA
v http://www.iana.org/assignments/port-numbers

Introduction 1-23
Hourglass design
q Question?
v Are TCP, UDP, and IP enough?
v What other functionality would applications
need?

Introduction 1-24
Hourglass design
q Security?
v IPsec/SSL/TLS

q Quality-of-service?
v RSVP, int-serv, diff-serv

q Reliable, out-of-order delivery service?


v SCTP

q Handling greedy sources?


q Accounting and pricing support?

Introduction 1-25
End-to-end principle and the
Hourglass design
q The good
v Basic network functionality allowed for
extremely quick adoption and deployment using
simple devices
q The bad
v New network features and functionality are
impossible to deploy, requiring widespread
adoption within the network
v IP Multicast, QoS

Introduction 1-26
Why did the Internet win?
q Packet switching over circuit switching
q End-to-end principle and “Hourglass” design
q Layering of functionality
q Distributed design, decentralized control
q Superior organizational process

Introduction 1-27
Layering
q Modular approach to network functionality
v Simplifies complex systems
• Each layer relies on services from layer below and
exports services to layer above
v Hides implementation
v Eases maintenance and updating of system
• Layer implementations can change without disturbing
other layers (black box)

Introduction 1-28
Layering
q Examples:
v Topology and physical configuration hidden by
network-layer routing
• Applications require no knowledge of routes
• New applications deployed without coordination with
network operators or operating system vendors

Application
Host-to-host connectivity
Link hardware

Introduction 1-29
Layering essential in Protocols
q Set of rules governing communication
between network elements (applications,
hosts, routers)
q Protocols specify:
v Interface to higher layers (API)
v Interface to peer
• Format and order of messages
• Actions taken on receipt of a message

Introduction 1-30
Layering: Internet protocols
q application: supporting network applications
v FTP, SMTP, HTTP
q transport: process-process data transfer
application
v TCP, UDP
q network: routing of datagrams from source transport
to destination
v IP network
q link: data transfer between neighboring
network elements link
v Ethernet, 802.11
q physical: bits “on the wire” physical

Introduction 1-31
source
message M application
Layers in action
segment Ht M transport
datagram Hn Ht M network
frame Hl Hn Ht M link
physical
Hl Hn Ht M link Hl Hn Ht M
physical

switch

destination Hn Ht M network Hn Ht M
M application Hl Hn Ht M link Hl Hn Ht M
Ht M transport physical
Hn Ht M network
Hl Hn Ht M link router
physical

Introduction 1-32
Why did the Internet win?
q Packet switching over circuit switching
q End-to-end principle and “Hourglass” design
q Layering of functionality
q Distributed design, decentralized control
q Superior organizational process

Introduction 1-33
Distributed design and control
q Requirements from DARPA
v Must survive a nuclear attack

q Reliability
v Intelligent aggregation of unreliable
components
v Alternate paths, adaptivity

q Distributed management & control of


networks
v Allows individual networks to independently
develop without large amounts of coordination
v Exceptions: TLDs and TLD servers, IP address
allocation (ICANN)
Introduction 1-34
Superior organizational process
q IAB/IETF process allowed for quick
specification, implementation, and
deployment of new standards
v Free and easy download of standards
v Rough consensus and running code
v 2 interoperable implementations
v Bake-offs
v http://www.ietf.org/

q ISO/OSI
v Comparison to IETF left as an exercise

Introduction 1-35
Internet history

Introduction 1-36
Internet History
1961-1972: Early packet-switching principles

q 1961: Kleinrock - queueing q 1972:


theory shows v ARPAnet public demonstration
effectiveness of packet-
v NCP (Network Control Protocol)
switching
first host-host protocol
q 1964: Baran - packet-
v first e-mail program
switching in early military
nets v ARPAnet has 15 nodes
q 1967: ARPAnet conceived
by Advanced Research
Projects Agency
q 1969: first ARPAnet node
operational

Introduction 1-37
Internet History
1972-1980: Internetworking, new and proprietary nets
q 1970’s: proprietary network Cerf and Kahn’s internetworking
architectures developed: principles:
DECnet, SNA, XNA v minimalism, autonomy - no
internal changes required
q 1974: Cerf and Kahn -
to interconnect networks
architecture for
v best effort service model
interconnecting networks
v stateless routers
q 1976: Ethernet at Xerox v decentralized control
PARC define today’s Internet
q 1979: ARPAnet has 200 nodes architecture

Introduction 1-38
Internet History
1980-1990: new protocols, a proliferation of networks

q 1983: deployment of q Late 1980s, Early


TCP/IP 1990s: new national
q 1983: smtp e-mail networks: Csnet,
protocol defined BITnet, NSFnet,
q 1983: DNS defined Minitel
for name-to-IP- v 100,000 hosts
connected to
address translation confederation of
q 1985: ftp protocol networks
defined
q 1988: TCP congestion
control
Introduction 1-39
Internet History
1990, 2000’s: commercialization, the Web, new apps

q Early 1990’s: ARPAnet Late 1990’s – 2000’s:


decommissioned
q more killer apps: instant
q 1991: NSF lifts restrictions on messaging, P2P file sharing
commercial use of NSFnet
q network security to
(decommissioned, 1995)
forefront
q early 1990s: Web
q est. 50 million host, 100
v hypertext [Bush 1945, Nelson million+ users
1960’s]
q backbone links running at
v HTML, HTTP: Berners-Lee Gbps
v 1994: Mosaic, later Netscape
q late 1990’s: commercialization of
the Web

Introduction 1-40
Internet History

2007:
q ~500 million hosts
q Voice, Video over IP
q P2P applications: BitTorrent
(file sharing) Skype (VoIP),
PPLive (video)
q more applications: YouTube,
gaming
q wireless, mobility

Introduction 1-41

Das könnte Ihnen auch gefallen