Anda di halaman 1dari 23

Basics of Communication

and the Internet


Circle Lecture Communication Systems
Winter Term 2004/2005

Prof. Dr. M. Zitterbart


Institute of Telematics
Dr.-Ing. Roland Bless

Outline
‰ Communication trends and scalability
‰ Basics of data communication
‰ How the Internet works
‰ Design Principles and threats for the Internet architecture

Communication Systems – Basics of communication and Internet – 1.1 04/05 www.tm.uka.de

Communication Trends
Mobile Communications
‰ Paradigm: anybody, anytime, anywhere
‰ Expected: more mobile phone subscribers than POTS subscribers
(Germany: already 48 Mio. at the end of 2000)

Technical Communications
‰ Today: communication between users
‰ Tomorrow: communication between machines, e.g.
‰ Production infrastructure: tele-metrics, tele-diagnosis, tele-operations
‰ Communications between vehicles:
‰ Home networks: sensors, security, appliances

IP-based Communications
‰ Internet Protocol IP as media independent access
‰ Voice-Over-IP technology is rolling out
‰ “All-IP” networks: Telcos will switch to IP for voice calls

Communication Systems – Basics of communication and Internet – 1.2 04/05 www.tm.uka.de


„Everything goes IP”

IP

IP

IP

Communication Systems – Basics of communication and Internet – 1.3 04/05 www.tm.uka.de

Internet Growth

Survey based on #hosts registered in DNS


#Hosts worldwide (Mio.)
300

250

200

150

100

50

0
91 92 93 94 95 96 97 98 99 00 01 02 03 04
Year

Communication Systems – Basics of communication and Internet – 1.4 04/05 www.tm.uka.de


Growth and Scalability
Constant change is presumably the only constant in the Internet
Internet survived the tremendous growth: it still works!
‰ One says: it is “scalable”
‰ What means scalability?
Scalability
A scalable system works even when there is tremendous growth (e.g., by
several orders of magnitude, i.e., over several scales) of certain system
parameters
Why important? Technological development shows often leaps in order of a
magnitude (c.f. Moore‘s Law, CPU, bandwidth, memory)
Example for no or bad scalability:
System
X(t) performance

t X(t)
Performance of a non-scalable system decreases (strongly) as certain
parameter values increase, possibly until the whole system fails
Communication Systems – Basics of communication and Internet – 1.5 04/05 www.tm.uka.de

Evolving Internet – important aspects


Past
‰ Data communication between research institutions
‰ Common goals
‰ Trust relationships between users
‰ Technically skilled users
‰ Consistent and coherent architecture
Presence
‰ Global infrastructure of the information society
‰ New interest groups and commercialization (ISPs, service providers)
‰ Loss of trust relationships
‰ Average consumers, technically unskilled
‰ Out of own interests, technologies and extensions are realized, which
z are used for short-time fulfillment of demand
z are largely done without architectural thinking
z are not consistent with the Internet architecture
z endanger the coherence of the internet

Communication Systems – Basics of communication and Internet – 1.6 04/05 www.tm.uka.de


Data Communications
‰ Communication (original meaning):
„Exchange of data between human communication partners.“

ÖEvery concrete communication is data communication


N.B.: Information is extracted from data by the process of interpretation

‰ Data communication
(more narrow definition in literature and habitual language use):
„Transmission of digital data between telecommunication devices“

‰ Communication (Usage of the term in this lecture):

„Data
„Data(tele)communication
(tele)communicationisisthe
thegeneric
genericterm
termfor
foreach
eachdata
data
exchange using immaterial media and greater distances
exchange using immaterial media and greater distances
between
betweenmenmenand/or
and/ormachines
machines
(abbreviated:
(abbreviated: Data communication==communication).“
Data communication communication).“
‰ immaterial media:
z Energy flows, usually electric currents, electromagnetic waves
z Opposite: material data transport (e.g. letters, shipping of disks)

Communication Systems – Basics of communication and Internet – 1.7 04/05 www.tm.uka.de

Basic model of telecommunication

sender service interface receiver


service
access point

message

medium

spatial distance

‰ Participants act as senders or receivers


‰ The service usage by participants occurs at a special service interface, using
a service access point
‰ Different service primitive types: Request, Indication, Response, Confirmation
‰ The Medium bridges the spatial distance

Communication Systems – Basics of communication and Internet – 1.8 04/05 www.tm.uka.de


What is a protocol?
A communication protocol describes a set of rules, according to which the
communication between two or more parties must be performed.

Communication protocols

e.g. discussion, conversation

Computer communication protocols


e.g. file transfer, electronic mail
ISO/OSI protocols
IPX DECnet
TCP/IP protocols

Ethernet AppleTalk WLAN

Communication Systems – Basics of communication and Internet – 1.9 04/05 www.tm.uka.de

Service and Protocol

Service User 1 Service User 2


Service
Service

Service Protocol Service


Provider 1 Provider 2

Communication Systems – Basics of communication and Internet – 1.10 04/05 www.tm.uka.de


A Model for Telecommunication Systems

Sender Receiver

telecommunication system

entity
entity nn layer n entity
entity nn

entity
entity n-1
n-1 layer n-1 entity
entity n-1
n-1

...
...

entity
entity 11 layer 1 entity
entity 11

Physical medium

‰ A layer offers a service to its upper layer


‰ The service is provided by the cooperation of the layer entities
according to a specified protocol

Communication Systems – Basics of communication and Internet – 1.11 04/05 www.tm.uka.de

ISO/OSI and Internet Model


ISO/OSI Basic Internet
Reference Model Reference Model
7 Application
6 Presentation Application
5 Session
4 Transport Transport
3 Network Internet
2 Data Link
Media Access
1 Physical

‰ ISO/OSI too complex, but OK as logical model


‰ Too restrictive (no cross-layer information exchange)
‰ Redundant functionality in different layers
‰ Too heavy-weight for simple network devices like printers, etc.

‰ Internet model similar, but simplified (esp. Application layer)

Communication Systems – Basics of communication and Internet – 1.12 04/05 www.tm.uka.de


Physical Layer
Tasks
‰ Accesses the physical medium directly (e.g. cable)
‰ Unsecured connection between systems
‰ Transport of unstructured bit sequences via a physical medium
‰ Comprises (among other things) physical link, conversion data ⇔ signals
Signal Transmission Modes S(t)

‰ Baseband Transmission
t
‰ Native and fully digital:
discrete signal levels, periodic and discrete transition intervals
‰ Maximum data rate for channel with bandwidth B according to
z Nyquist: rmax [bit/s] =2 B log2 n, (n=number of discrete levels, noise-less channel)
z Shannon: rmax [bit/s] = B log2 (1 + S/N) (noisy channel, S/N=Signal-to-noise ratio)
‰ Broadband Transmission
‰ Modulation (amplitude, frequency, phase or combination thereof)
S(t)
‰ Modem (modulator/demodulator) required

t
Communication Systems – Basics of communication and Internet – 1.13 04/05 www.tm.uka.de

Medium-Access/Data Link Layer


Tasks
‰ Structuring the data stream
‰ Synchronization, Framing, Code Transparency
‰ Protection against errors and loss
‰ Use of checksum to detect bit errors (e.g., CRC: Cyclic Redundancy Check)
‰ Reliable link layers use sequence numbers, timers and acknowledgments to
detect loss of data packets and to recover by automatic retransmission
‰ Flow control
‰ Media access control in case of shared media
Network Access
‰ Local Area Networks, e.g., Ethernet, Token-Ring, Token-Bus, Wireless
LANs, ....
‰ Ethernet-Frame
Preambel
Preambel StartDel
StartDel DestAddr
DestAddr SrcAddr
SrcAddr Length
Length Data
Data PAD
PAD FCS
FCS
56 bit (8 bit) (16/48 bit) (16/48 bit) (16 bit) (≤12.000 bit) (0-368 bit) (32bit)
56 bit (8 bit) (16/48 bit) (16/48 bit) (16 bit) (≤12.000 bit) (0-368 bit) (32 bit)
‰ Metropolitan Area and Wide-Area Networks: Modems, Fiber, DSL, ...

Communication Systems – Basics of communication and Internet – 1.14 04/05 www.tm.uka.de


Network Layer
Tasks
‰ Concatenation of point-to-point
connections to end-system connections
‰ Uniform addressing of nodes End-system A End-system B
Application-
‰ Address mapping to oriented
data link layer addresses Layers

‰ Transmission quality possibly Transport TCP


TCP
Intermediate System
TCP
TCP
Layer
selectable
IP
IP IP
IP IP
IP
Network
‰ Routing Layer
‰ Flow control, congestion control Media Access IEEE
IEEE 802.3
802.3 IEEE
IEEE 802.3
802.3 IEEEIEEE 802.5
802.5 IEEE
IEEE 802.5
802.5

Physical Medium Physical Medium


Switching concepts
‰ Circuit Switching (Classical telephony, e.g. ISDN)
‰ Packet Switching (Internet)
‰ Virtual Connections (ATM: Asynchronous Transfer Mode)
‰ Message Relaying
Communication Systems – Basics of communication and Internet – 1.15 04/05 www.tm.uka.de

Network Layer: Internet Protocol


‰ IP layer enables
‰ Bigger network E-Mail, WWW, Telephony ....

‰ Global addressing
‰ Hide network details and changes SMTP, HTTP, RTP, BEEP, ...
from end-to-end protocols
‰ A single protocol (Hourglass Model) UDP, TCP,
SCTP, ...
‰ maximizes interoperability
‰ minimizes the number of service interfaces IP
‰ Lean protocol
‰ Requires minimal common network functionality Ethernet,
in order to maximize the number PPP, ...
of usable networks
‰ End-to-End principle CSMA, CDMA, Asynch., SDH, ...
‰ Robustness by stateless operation
Copper, Glass Fibre, Radio, ...

‰ See also:
http://www.iab.org/Documents/hourglass-london-ietf.pdf
Communication Systems – Basics of communication and Internet – 1.16 04/05 www.tm.uka.de
Routing in the Internet
Problem
‰ How are data packets forwarded in the Internet?
Method
‰ Routing table gives information about the next hop
‰ The protocol IP Transport layer: UDP, TCP
(Internet Protocol) conducts the Protocol
ProtocolIPIP
• •Addressing
forwarding of data Routing protocols
Routing protocols Addressing
• Datagram format
• •RIP,
RIP,OSPF,
OSPF,BGP
BGP • Datagram format
‰ Datagram protocol •„Packet
•„Packethandling“
handling“
z connectionless Protocol
ProtocolICMP
ICMP
Address
Addressresolution • •error
resolution errorreports
reports
z unreliable • •ARP,
ARP,RARP
RARP Routing • •Signalling
Signallingbetween
between
table routers
z segmentation and routers
reassembly Data link layer

‰ Uses Internet addressing Physical layer

‰ Uses further protocols like


z ICMP (Internet Control Message Protocol)
z ARP (Address Resolution Protocol)
z IGMP (Internet Group Management Protocol)
Communication Systems – Basics of communication and Internet – 1.17 04/05 www.tm.uka.de

Format of an IPv4 data unit


012 345 67
According to PPPDTR 00
RFC 791
Version (4) Header Length (4) (obsolete)
Type of Service (8) reserved
Precedence
Total Length (16) 111 Network Control
110 Internetwork Control
Identifier (16)
101 CRITIC/ECP
Flags (3) Fragment Offset (13) 100 Flash Override
011 Flash Delay: 0 normal
Time to Live (8) 010 Immediate 1 low
Protocol (8) 001 Priority Throughput: 0 normal
000 Routine 1 high
Header Checksum (16)
Reliability: 0 normal
Source Address (32) 1 high
NEW:
Destination Address (32) 0 1 2 3 4 5 6 7
Options and Padding (variable) R
Data (variable) DS Field ECN
Differentiated Services Explicit
Congestion
Notification

Communication Systems – Basics of communication and Internet – 1.18 04/05 www.tm.uka.de


IP addresses
Structure of IPv4 addresses
network
networkpart
part local
localpart
part

network
networkpart
part subnet
subnetpart
part end
endsystem
system
Subnet masks mark the area of the IP address describing the network and the sub-
network. This area is marked as ones („1“) in the binary form of the subnet mask.
Example
IP address: 129. 13. 3. 64
Subnet mask: 255. 255. 255. 0 =
1111 1111 1111 1111 1111 1111 0000 0000

Network written in prefix notation: 129.13.3.0/24


Globally visible network is only 129.13.0.0/16 (formerly Class B network)
Network: 129. 13.
Subnet: 3.
End system: 64
‰ If the subnet mask only covers the network part, there is no subnet part
(e.g. subnet mask 255.255.0.0 in case of class B)
‰ Note: Systems attached to several networks (e.g. routers), have several, network-specific
IP addresses!
Communication Systems – Basics of communication and Internet – 1.19 04/05 www.tm.uka.de

Mapping of IP and MAC addresses


‰ If (Destination IP address AND Subnet mask)
= (Own IP address AND Subnet mask)
Æ Receiver is in the same IP subnet! So I can use a link layer connection...
‰ Problem:
Which MAC address does the next system on the route to the target have?
Scheme

Application Application

Connect with
TCP TCP TCP
12 . 0 . 0 . 21

IP 12 . 0 . 0 . 34 12 . 0 . 0 . 21 IP

MAC MAC
08002B90102456 ????????????????

Internet

Communication Systems – Basics of communication and Internet – 1.20 04/05 www.tm.uka.de


Forwarding in an IP router
End system A End system B
129.13.3.108 IP-Router 1 145.5.9.27
129. 132.
13. 2.
3. 2.
60 3
MAC-A MAC-B
132. 145.
2. 5.
2. 9.
7 19
Routing table (Routing Information Base) IP-Router 2
‰ Constructed by routing protocols: contains several alternative routes to the destination
Forwarding table (Forwarding Information Base)
‰ Only selected/active routes: IP address of next hop and identification of the interface used
Address resolution table
‰ Built by ARP: MAC address of the next system for the IP address of the end system
Example
‰ Destination: end system B; Source: end system A
‰ Data packet on the way from router 1 to router 2:
z MAC addresses: MAC address IP-Router 2 (dest) and MAC address IP-Router 1
(source)
z IP addresses: end system B (destination), end system A (source)
Communication Systems – Basics of communication and Internet – 1.21 04/05 www.tm.uka.de

Forwarding in an IP router
Network scenario with router
End system A
End system B
129.13.3.108
145.5.9.27

Router 1 Router 2
129.13.3.60 132.2.2.7
132.2.2.3 145.5.9.19
M AC-A
If A If B If A If B M AC-B

129.13. 145.5. ... M AC- M A C-


...
3.108 9.27 R1-B R2-A

IP addresses MAC addresses


Router functions
‰ Determine the IP address of the subsequent system (Next Hop)
‰ Simple routers have often only a routing table for their subnets and a default route for all
other destinations
‰ Mapping of this IP address to the connection point address (MAC address)
‰ Sending the IP data unit to the next hop on the corresponding interface via layer 2
‰ IP addresses (Source/Destination) in the IP packet remain unchanged!
Communication Systems – Basics of communication and Internet – 1.22 04/05 www.tm.uka.de
Routing in the Internet
Network layer protocols Protocols in an IP router
‰ IP(Internet Protocol)
‰ ARP (Address Resolution Protocol)
‰ RARP (Reverse ARP)
‰ ICMP (Internet Control Message Protocol)
‰ IGMP (Internet Group Management
BGP RIP SNMP
Protocol)
‰ SNAP (Subnetwork Access Protocol) EGP /
TCP UDP IGP
OSPF
Routing protocols
ICMP IGMP
‰ RIP(Routing Information Protocol) Internet Protokoll
‰ BGP (Border Gateway Protocol)
‰ EGP (External Gateway Protocol) ARP RARP
‰ OSPF (Open Shortest Path First) SNAP
Network management LLC-1
‰ SNMP (Systems Network Management
Protocol)
Transport protocols:
‰ UDP (Universal Datagram Protocol)
‰ TCP (Transmission Control Protocol)

Communication Systems – Basics of communication and Internet – 1.23 04/05 www.tm.uka.de

Routing Hierarchy: View from 10,000m


AS 120
AS 100 AS 111 AS 112

AS 121
AS 110 AS 114

AS 101 AS 113
AS 122

Splitting networks into „Autonomous Systems“ (AS)


‰ Otherwise entries in routing tables and amount of exchanged routing
information not scalable
‰ Routers within AS have usually only detailed routing information about own AS
‰ There is at least one designated router that acts as interface to other ASes
‰ Advantages
z Scalability
– Internal routing table size depends on size of AS
– Changes within AS are usually only propagated within the AS if external connectivity is not
affected
z Autonomy
– Internet = Network of networks
– Routing is controlled by own organization
» Unique routing strategy within own system
» Internal routing protocols can vary between ASes
Communication Systems – Basics of communication and Internet – 1.24 04/05 www.tm.uka.de
AS and prefix number growth
‰ Each AS has a unique number
(currently 16 bit, extension to 32 bit
planned)
‰ Currently (Oct. 2004) ~18,000 ASes

‰ Currently (Oct. 2004) ~180,000 different


IPv4 network prefixes (= best routes)

Source: http://bgp.potaroo.net

Communication Systems – Basics of communication and Internet – 1.25 04/05 www.tm.uka.de

Routing und Autonomous Systems


Autonomous Systems are interconnected
ISP
ISP
‰ Stub-AS
z Small companies
z Connection to exactly one provider Stub-AS
‰ Multihomed Stub-AS
z Big companies ISP
ISPAA ISP
ISPBB
z Connection to several providers (resilience)
z No transit traffic Transit-AS Stub-AS
‰ Transit AS
ISP
ISPAA ISP
ISPBB
z Provider

Two different levels of routing


‰ Intra-AS
z Administrator is responsible for selecting a routing protocol
‰ Inter-AS
z Uniform standards (BGP)
z AS announces which networks (prefixes) can be reached through it
– Own networks located/homed in this AS (Origin AS)
– Networks in other foreign ASes (then the AS is willing to be transit AS for these destinations)

Communication Systems – Basics of communication and Internet – 1.26 04/05 www.tm.uka.de


Why Intra- and Inter-AS routing protocols?
Policy
‰ Political question: which transit traffic is allowed to traverse the AS?
‰ Inter-AS: policies are selected by the provider
‰ Intra-AS: one organization, few policies necessary

Scalability
‰ Inter-AS: further abstraction level;
Size of routing tables and number of updates can be reduced, as failures within
one AS can mostly remain hidden
‰ Intra-AS: higher stability

Performance
‰ Inter-AS: Policies are necessary and more important than performance metrics
‰ Intra-AS: Concentration on performance metrics

Communication Systems – Basics of communication and Internet – 1.27 04/05 www.tm.uka.de

Intra-AS Routing
Well-known protocols for Intra-AS routing are
‰ RIP (Routing Information Protocol) Æ Distance Vector Protocol
‰ OSPF (Open Shortest Path First) Æ Link State Protocol
‰ IS-IS (Intra-Domain Intermediate System to Intermediate System Routing
Protocol) Æ Link State Protocol
z originally ISO/OSI routing protocol
z used for IP by big providers
‰ EIGRP (Enhanced Interior Gateway Routing Protocol)
z CISCO proprietary

Intra-AS routing protocols are often called Interior Gateway Protocols (IGP)
OSPF:
Vertex=Node
‰ Connectivity and link states are flooded through the network (router/
‰ Every router has the same view of the network subnet)
‰ Network is mapped to Graph (V,E)
‰ Calculates shortest paths
with Dijkstra’s algorithm Edge=Link

Communication Systems – Basics of communication and Internet – 1.28 04/05 www.tm.uka.de


OSPF hierarchy in an Autonomous System

interior Router
N1
border router
R1 R13
R3 (ASBR) virtual
N2 (ABR) connection

R2 Routing Area
(OSPF Area)
R4 R12
(ABR) (BBR) N: Network

Autonomous System R: Router


N3 ABR: Area Border Router
ASBR/
ABR ASBR: AS Boundary Router
R5 R9
BBR: Backbone Router
R7 ASBR/
(ABR) R8 ABR
R6
R11

N4 R10
ASBR/
ABR

Communication Systems – Basics of communication and Internet – 1.29 04/05 www.tm.uka.de

Inter-AS Routing: Exterior BGP (EBGP)


‰ Exterior BGP is used between the BGP routers (also called BGP
speakers) connecting two ASes
‰ Path Vector protocol (AS path)
‰ Learn all destination prefixes that can be reached through the other AS
‰ An AS can prevent to receive traffic for certain destination by not
announcing any route to it (i.e., policy by route filtering)
‰ These BGP routers should be directly connected
‰ Internal information will NEVER be forwarded directly to other BGP
speakers

AS 1 AS 2

BGP Speaker

Communication Systems – Basics of communication and Internet – 1.30 04/05 www.tm.uka.de


Example for BGP topology
I want to send data to AS122!
Which route should I use? AS 120
AS 112

120.0.0.0/8
100.0.0.0/8 111.0.0.0/8 112.0.0.0/8 AS 121

AS 100 AS 111

AS 101 121.0.0.0/8
110.0.0.0/8 114.0.0.0/8

AS 110 AS 114
101.0.0.0/8 AS 113 113.0.0.0/8
AS 122 122.0.0.0/8

Routing table AS100


Network Next Hop Metric LocPrf Weight Path

Routing table AS110
Network Next Hop Metric LocPrf Weight Path
*> Routing
121.0.0.0 10.1.1.110
table AS114 0 110 114 121 i
* 121.0.0.0 10.1.1.111 0 111 112 114 121 i
*> 122.0.0.0 10.1.1.110 0 110 114 122 i
* Network Next Hop
10.1.1.113 Metric LocPrf Weight0 Path
113 114 121 i
*>*> 112.0.0.0 10.1.1.112
10.1.1.114 0 0 112
0 114 121i i
*>* 122.0.0.0 10.1.1.110
10.1.1.114 0 110
0 114 122111i 112 i
*… 10.1.1.111 0 111 112 114 122 i
* *> 122.0.0.0 10.1.1.122
10.1.1.113 0 0 0113
122114
i 122 i

Communication Systems – Basics of communication and Internet – 1.31 04/05 www.tm.uka.de

Routing in "Default-Free-Zones"
Two modes of operation of BGP (same protocol, IBGP
but different rules) for the distribution EBGP
of routing information
Between two AS: with EBGP (External BGP)
Within one AS: with IBGP (Internal BGP)
Internal full mesh of
TCP connections necessary
No distribution of routes learnt with EBGP
IBGP to IBGP neighbors IBGP

EBGP AS X
AS Y

EBGP
IBGP EBGP

Full mesh unsuitable for big AS,


EBGP possible solution lies in the implementation of
Route reflectors (dedicated routers as
peering points)
Confederations (private sub-AS)

Communication Systems – Basics of communication and Internet – 1.32 04/05 www.tm.uka.de


Interior BGP (IBGP)
‰ BGP routers within one AS are connected with IBGP
‰ IBGP routers have to be fully meshed
z To learn routes for all external prefixes
z They inform about new networks (e.g. LANs)
z They do not propagate internal prefixes outwards
‰ No direct physical connections (but logical connections) between routers
necessary
‰ Each IBGP router must be able to communicate with each other IBGP
router
‰ IBGP messages are never forwarded to other BGP routers (to prevent
loops)

Communication Systems – Basics of communication and Internet – 1.33 04/05 www.tm.uka.de

Transport Layer
Tasks
‰ End-to-end service
‰ application-based addressing (Ports)
‰ reliable/unreliable
‰ Reliable protocol
‰ Error and loss detection
‰ Retransmission
‰ Segmentation/Reassembly
‰ Flow control
‰ Congestion control

Examples
‰ TCP (Transmission Control Protocol)
‰ UDP (User Datagram Protocol)
‰ SCTP (Stream Control Transmission Protocol)
‰ DCCP (Datagram Congestion Control Protocol)

Communication Systems – Basics of communication and Internet – 1.34 04/05 www.tm.uka.de


Application Layer
‰ Protocols depend on the particular application
‰ This is also end-to-end
Examples of protocols above the transport layer:
‰ telnet (Remote Login)
‰ SSH (Secure Shell, secure replacement for telnet)
‰ FTP (File Transfer Protocol)
‰ HTTP (Hypertext Transfer Protocol, HTML/Web Content Transport,
Server/Client)
‰ BEEP (Blocks Extensible Exchange Protocol, Peer-to-Peer, many features)
‰ SSL/TLS (Transport Layer Security)
‰ SMTP (Mail Transport)
‰ DNS (Domain Name System)
‰ RTP (Streaming)
‰ Routing Protocols (OSPF, BGP, ...)
‰ ...many more...

Communication Systems – Basics of communication and Internet – 1.35 04/05 www.tm.uka.de

Internet architecture: Design goals


Paper by D. Clark “The Design Philosophy of the DARPA Internet Protocols”
(SIGCOMM '88) names:
‰ Fundamental goal: Internetworking (Connection of existing networks)
‰ Further goals (ordered by their importance):
‰ Robustness: sustain internet communication despite failure of networks and
routers
‰ Support of multiple types of communication services
‰ Heterogeneity: Accommodation of a variety of networks
‰ Distributed resource management
‰ Cost effectiveness
‰ Host attachment with a low level of effort
‰ Resources used must be accountable
‰ Robustness against failures
‰ „Fate-Sharing“: acceptable to loose the state information associated with an
entity if, at the same time, the entity itself is lost
‰ Do not store state in the network, but in the end systems instead
‰ Datagram concept as a consequence
Communication Systems – Basics of communication and Internet – 1.36 04/05 www.tm.uka.de
Design principles: End-to-End Argument
Decisions necessary in system design
‰ Which functionality is needed?
‰ Where should certain functions be placed?
‰ In the end systems or applications?
‰ In the network?

Important design principle


(explicitly expressed as recently as 1981 by Saltzer, Reed and Clark)

The End-to-End-Argument (E2E argument):


„The function in question can completely and correctly be
implemented only with the knowledge and help of the application standing
at the end points of the communication system. Therefore, providing that
questioned function as a feature of the communication system itself is not
possible. (Sometimes an incomplete version of the function provided by the
communication system may be useful as a performance enhancement.)“
Communication Systems – Basics of communication and Internet – 1.37 04/05 www.tm.uka.de

Discussion End-to-End-Argument
‰ This means especially:
specific functionality of the application layer usually can and should
preferably not be placed in the network itself
‰ Minimality principle:
‰ Avoid integrating more than the essential and necessary functionality into the
network
‰ Keep unnecessary functionality out of the networkÆ Keep it simple
‰ Not a strict law, rather a guideline

Further goals and consequences of the E2E argument:


‰ Protection of innovation
‰ Simple to add new services
‰ Hard to change the infrastructure (see introduction of Multicast, IPv6, ECN, etc.)
‰ Reliability and robustness
‰ against failure and malfunction of end systems and network components
‰ If network components have to store state, the probability of connection failures
grows with increasing network size
Communication Systems – Basics of communication and Internet – 1.38 04/05 www.tm.uka.de
Consequences End-to-End-Argument
Examples:
‰ Reliable file transfer
Possible sources of error:
‰ Read errors in the end system
‰ Software errors during copying or buffering of data by the file system or file
transfer program
‰ Hardware errors during these processes in CPU, memory, bus, etc.
‰ Loss, bit errors or duplicates in the communication system
‰ Crash/Failure of end systems (sender or receiver) during or after transfer
‰ Reliability of the communication system does not eliminate all errors
‰ Division of TCP/IP into TCP and IP in the late 70’s
‰ End-to-End security
‰ Suppression of duplicates (e.g. caused by the application itself)

Communication Systems – Basics of communication and Internet – 1.39 04/05 www.tm.uka.de

Internet Architecture: Principles


RFC 1958: „Architectural Principles of the Internet“
‰ Independence of the Internet Protocol of the medium and of hardware
addressing
‰ If states have to be stored (e.g. routes, QoS-guarantees, Header
Compression, ...), they should be “self-healing”
‰ Adaptive procedures and protocols for deriving and maintaining states
‰ „Soft-State“ concept: State is periodically renewed („refreshed“)
‰ Reduction of state information to a minimum
‰ Manually configured states should be reduced to an absolute minimum

Communication Systems – Basics of communication and Internet – 1.40 04/05 www.tm.uka.de


Further design aspects
RFC 3426 „General Architectural and Policy Considerations“
(Internet Architecture Board)
‰ basic issues concerning protocol and system design
‰ no guidelines, no checklist
‰ Discussion and explanation on the basis of numerous
case studies (e.g. ECN)

RFC 1122 „Requirements for Internet Hosts -- Communication Layers“


‰ Good documentation and discussion of design decisions
‰ Robustness principle (Jon Postel, see also http://www.postel.org):
“Be liberal in what you accept, and conservative in what you send”
‰ Software should be able to react appropriately to every error – even if it
is highly unlikely
‰ Incoming packet can contain any combination of faults and attributes
‰ Assumption of intended/malicious generation of such packets

Communication Systems – Basics of communication and Internet – 1.41 04/05 www.tm.uka.de

Trends opposed to the E2E principle (1)


Many aspects have changed since the outset of the internet
„Threats“ to the End-to-End-Argument? [RFC 3724]
‰ Loss of trust between end systems
Æ Introduction of security technologies
Middlebox
‰ Middleboxes (Proxies/NATs/Firewalls/Caches/...)
Æ Break of the End-to-End Principle (esp. security mechanisms)
‰ New service models: Quality of service becomes part of the service
(Streaming A/V) Æ Servers are distributed and placed closer to the user
(e.g. Akamai, Realnetworks...)
‰ New parties involved: Internet Service Provider, Administrators of company
networks, governments
Æ Restriction of services, interest of interposing (e.g. as a Trusted Third
Party or for eavesdropping/taxation/censorship...)
‰ Technically uninterested users
Æ Context and configuration information is placed in the network in order to
disburden the user

Communication Systems – Basics of communication and Internet – 1.42 04/05 www.tm.uka.de


Trends opposed to the E2E principle (2)
‰ Example: negative effects by security technologies
‰ Elimination of PATH-MTU-Discovery mechanisms by rigorous filtering
of ICMP packets
‰ Filtering of packets with their ToS-Bits set prevents the usage of Explicit
Congestion Notification
‰ Limitation of accessibility and available services by private addressing
in “Intranets”

‰ Possible procedure for future mechanisms which seem to infringe upon the
End-to-End principle:
Split E2E-Argument into the components
‰ Protection of innovation
z Introduction of new mechanisms is easier in end systems
‰ Reliability/Robustness and trust
z add security, where necessary

Communication Systems – Basics of communication and Internet – 1.43 04/05 www.tm.uka.de

Loss of internet transparency


Internet Transparency [RFC 2775]:
‰ original concept of a single universal logical addressing scheme
‰ Mechanisms which allow packets to flow essentially unchanged from
source to destination
Loss of transparency by:
‰ Intranets („Security“, Restriction of applications and address transparency,
network administrator has control)
‰ Dynamic addresses (SLIP/PPP, DHCP)
‰ Firewalls (Restriction of services and accessibility)
‰ SOCKS/Application Level Gateways
‰ Private addresses (not unique, restriction of accessibility and global
communication)
‰ Network Address Translators (NATs)
‰ Application Level Gateways, Proxies, Caches
‰ Voluntary isolation (e.g. WAP-Proxies) and partner networks
‰ Split-DNS
‰ Tricks for load balancing

Communication Systems – Basics of communication and Internet – 1.44 04/05 www.tm.uka.de


Conclusions
Today we have networks everywhere, and, they are a critical part of the IT
infrastructure

Most network systems and architectures use Internet protocols

The Internet is a very scalable system


‰ Accommodated the tremendous growth in the past
‰ Thanks to the wise design decisions and architectural principles
‰ But for how long will its success continue?
‰ Requirements for more connectivity, machine-to-machine communication
etc. lead to the use of IPv6
‰ Current Inter-Domain routing scheme will probably fail to cope with growth
of the next two decades...
‰ Tussle and conflicts in the Internet caused by parties with different interests

Communication Systems – Basics of communication and Internet – 1.45 04/05 www.tm.uka.de

Anda mungkin juga menyukai