Computer Networks

Basics

Prof. Dr. Oliver Hahm

2023-10-20

Historical background

The ARPANET

  • 1957: Foundation of the Advanced Research Projects Agency (ARPA) by the US Dept of Defense (DoD) in response to Sputnik
  • 1962: The idea of the ‘Internet’ as ‘tool to create critical mass of intellectual resources’ (Licklider, Taylor)
  • 1967: Plan for the ARPANET was published
    Main architects: Vinton Cerf, Bob Kahn
  • 1969: First Request for Comments (RFC) and first functioning network, rented 50 kBit/sec lines, Interface Message Processors by BBN


Graphic by courtesy of Prof. Dr. Roland Kaiser, Hochschule RheinMain

First Internet Protocols

  • 1972: First public demo (remote login)
    using the Network Control Protocol (NCP)
    main use: terminal sessions, file transfer, Electronic Mail

  • 1974: Basics of TCP/IP written on paper by Cerf/Kahn (IP=Internet Protocol, TCP=Transmission Control Protocol), standardization in the following years

  • 1982: Transition towards IP version 4 (IPv4) 

  • from 1983:: Dissemination of TCP/IP due to Berkeley UNIX 4.2 BSD, source code publicly available

 

Standardization

  • 1986: The Internet Engineering Task Force (IETF) is founded as an open standardization organization
  • 1989: Foundation of RIPE (Réseaux IP Européens) as a forum for administrative and technical coordination of Internet development
  • 1990: Proposal of a hypertext project at CERN in Geneva by Tim Berners-Lee and Robert Cailliau: cradle of the world wide web
  • 1995: The specification of IPv6 (as a successor of IPv4) is published by the IETF

 

Global Success

  • 1996: First search engines with a site-scoring algorithm, e.g., Google search
  • 1998: Start of the dot-com boom
  • 2004: Start of Web 2.0 brought up blogs and RSS as well as services like Facebook or Twitter
  • 2007: Apple’s iPhone and Android started the “Mobile Revolution”
  • 2008: Rise of the Internet of Things (IoT)

 

Internet growth

  • Amount of AS (Autonomous Systems, admin. routing domain)

    • Doubling every five years (currently, more than 100,000)

    • Stable core

    • Major growth at the fringe

  • Traffic rate

    • Growth rate of about 26% per year estimated
  • Users

    • 2021: two third of the world population is “online” 

    • More than doubled during the last ten years 

    • Strongest growth outside the EU, Japan, and USA 

 

Components and Terms

Why do we need Computer Networks?

Please go to the survey at
https://pingo.coactum.de/137261

  • Which network services or applications do you know?
  • What are the main tasks of the network?

Purpose of Computer Networks

The general task of a computer network is to enable communication among the participants.

  • Resource sharing
    \(\Rightarrow\) assign different tasks to different computers
    \(\Rightarrow\) avoid bottlenecks
  • Resource pooling
    \(\Rightarrow\) combine the resources and functionalities of multiple machines
  • Resource balancing
    \(\Rightarrow\) increase the availability of the services by redundancy

Required Components to set up a Computer Network

What do we need for a computer network?

  • For setting up and running a computer network, these three components are required:
  1. \(\geq 2\) computers with network services running
  1. Transmission medium to send and receive data
  1. Network protocols
  • The devices are intended to communicate with each other or access shared resources
  • A network service provides a service for communication or shared resources usage
  • Computers in a network are called hosts
  • Some sort of a wire (e.g., copper or fiber-optic cables)
  • The air might serve as medium as well \(\rightarrow\) wireless data transmission
  • Rules that specify, how computers can communicate

Network Services

  • A network service provides resources to other devices in the network

  • Distinguished by their role:

    Server

    Provides a network service

    Client

    Uses (consumes) a network service

  • If each communication partner is server and client both,
    the participants are called peers
    (\(\Longrightarrow\) Peer-to-Peer networks)

  • The terms server, client, and peer typically refer only to network services and not to hardware
  • Almost any computer that acts as a server will also run client applications

Transmission Media

Different transmission media exists to setup a computer network.

  1. Guided transmission media
  2. Wireless transmission
  • Copper cable: Data is transferred as electrical impulses
  • Fiber-optic cable: Data is transferred as light impulses
  • Wireless transmission can be realized directed and undirected
  • Directed transmission can base on the following technologies:
    • Radio technology: Electromagnetic waves in the radio frequency spectrum (radio
      waves) (e.g., directed WLAN and satellite Internet access)
    • Infrared: Electromagnetic waves in the spectral range (e.g., IrDA)
    • Laser: Data is transferred as light impulses via Laser Bridge
  • Undirected wireless transmission is mostly based on radio technology (e.g., WLAN,
    cellular networks, terrestrial broadcasting and satellite broadcasting) or sonar

Protocols

A protocol is the set of all previously made agreements between communication partners, e.g.,

  • Rules for connection establishment and termination
  • Method of synchronization between sender and receiver (if any)
  • Measures for the detection and treatment of transmission errors
  • Definition of valid messages (vocabulary)
  • Format and encoding of messages
  • Protocols specify…
    • the syntax (= format of valid messages)
    • the semantics (= vocabulary and meaning of valid messages)

 

Computer Networks distinguished by their Dimension (1/2)

  • Depending on the dimension, different groups of computer networks are distinguished
  • Personal Area Network (PAN) or Body Area Network (BAN)
    • Network of small mobile devices, such as smart phones
    • Dimension: Few meters
    • Technologies: USB, FireWire, WLAN, Bluetooth, IrDA
  • Local Area Network (LAN)
    • Local network
    • Range covers an apartment, building, company site or university campus
    • Dimension: 500-1000 m
    • Technologies: Ethernet, Wireless LAN (WLAN)

Computer Networks distinguished by their Dimension (2/2)

  • Metropolitan Area Network (MAN)
    • Connects LANs
    • Range covers a city or agglomeration area
    • Dimension: 100 km
    • Technologies: Fiber-optic cables, WiMAX (IEEE 802.16)
      • Fiber-optic cables are used because of lesser attenuation (signal weakening) and higher data transmission rates
  • Wide Area Network (WAN)
    • Connects several networks
    • Range covers a large geographic area inside a country or continent
    • Dimension: 1000 km
    • Technologies: Ethernet (10 Gbit/s), Asynchronous Transfer Mode (ATM)

Communication Modes

  • Synchronous (“Rendez-Vous”)
    • Sender and receiver needs to be present at the same time
    • May require to wait for the other side to become ready
    • For example, phone calls or video conference
  • Asynchronous
    • Sender and receiver may act independently from each other
    • Requires buffering
    • For example, instant messaging or E-Mail

Unicast and Broadcast

Unicast

One-to-one communication, i.e., one host sends information to exactly one other host

Broadcast

One-to-all communication, i.e., one host sends information to all other hosts in the network

Group Communication: Multicast and Anycast

Multicast

Group communication, i.e., one host sends information to all hosts in a given group

Anycast

One-to-any communication, i.e., one hosts sends information to one host in a given group

Connection-Orientation

Network services may operate connection-oriented or connectionless.

connection-oriented

the service operates stateful

  • comprises three phases: connection establishment, data transfer, and connection termination
  • a virtual path between the involved hosts is established
  • sequent data is exchanged between the same hosts
  • typically used for reliable services
connectionless

the service operates stateless

  • no path between the involved hosts is established
  • typically used for low latency services

Directional Dependence (Anisotropy) of Data Transmission

Given a communication channel with two (or more) endpoints:

  • Simplex
    • Only one side of the channel can send data \(\rightarrow\) the channel can be used in only one direction
    • Examples: Radio, TV, Pager
  • Duplex (Full-duplex)
    • Both sides of the channel are allowed to send \(\rightarrow\) the channel can be used in both directions simultaneously
    • Examples: Phone, Networks with twisted pair cables because they provide separate wires for send and receive
  • Half-duplex
    • Both sides of the channel can send, but not simultaneously \(\rightarrow\) the channel can only be used in one direction at a time
    • Examples:
      • Networks with fiber-optic cables or coaxial cables, because there exists just a single line to sending and receiving
      • Wireless networks with just a single channel

Bandwidth, Throughput and Goodput

Main factors, influencing the performance of a computer network:

  • Bandwidth (\(\rightarrow\) throughput)
  • Latency (delay)
  • The bandwidth specifies how many bits can be transmitted within a period via the network
    • If a network has a bandwidth of 1 Mbit/s, one million bits can be transmitted per second in the ideal case
      • Thus, a bit has a width of 1 \(\mu\)s
    • Throughput is the actual achieved data rate (\(\Rightarrow\) the bandwidth defines its upper bound)
    • Goodput is the actual rate of data the user benefits from

Latency

The latency of a network is the time, a message needs to travel from one end of the network to the most distant end

Latency = Propagation delay + Transmission delay + Waiting time

\[\mbox{Propagation delay} = \frac{\mbox{Distance}}{\mbox{Speed of light}*\mbox{Velocity factor}}\]

  • Distance: Length of the network connection
  • Speed of light: \(299,792,458\) m/s
  • Velocity factor: Vacuum = 1, twisted pair cables = 0.6, optical fiber = 0.67, coaxial cables = 0.77

\[\mbox{Transmission delay} = \frac{\mbox{Message size}}{\mbox{Bandwidth}}\]

Transmission delay = 0, if the message consists only of a single bit

  • Waiting times are caused by network devices (e.g., Switches)

    • They need to cache received data first before forwarding it

    • Waiting time = 0, if the network connection between sender and destination is just a single line or a single channel

Source: Larry L. Peterson, Bruce S. Davie. Computernetzwerke. dpunkt (2008)

Bandwidth-Delay Product

  • Calculates the volume of a network connection
    • Signals cannot be transmitted with infinite speed via the transmission media
      • The propagation speed is in any event limited by the speed of light and it depends on the velocity factor of the transmission medium
    • The product of bandwidth and delay (latency) corresponds to the maximum number of bits that can reside inside the line between sender and receiver
  • Example: A network with 100 Mbit/s bandwidth, and 10 ms latency

\[100,000,000\,\mbox{Bits/s} \times 0.01\,\mbox{s} = 1,000,000\,\mbox{Bits}\]

  • There are a maximum number of \(1,000,000\) Bits inside the network line
    • This is equivalent to \(125,000\) Bytes (approx. \(123\) kB) :::

How does a Computer Network work?

You need information about someone/something:

  • What do you do?

  • Which problems are to solve?

The Big Picture

Let’s create it together

Reference Models

Excursion: Software Engineering

Which steps are necessary to realize an IT project?

Recap

Let’s go again to the survey at
https://pingo.coactum.de/137261

  • Which components do we require for a computer network?
  • Name some properties to characterize a computer network
  • Which characteristics do apply for a phone call?

Reference Models

  • Reference models are used to describe computer networks independently of concrete technologies
  • Such a reference model consists of several layers
  • Each layer addresses a particular aspect of communication and offers interfaces to the neighboring layer
  • Each layer defines their own protocols that define syntax and semantics of parts of a transmitted message (e.g., header and trailer)
  • These message parts are encapsulated
  • Because each layer is complete in itself, single protocols can be modified or replaced without affecting all aspects of communication
  • The most popular reference models are…
    • the TCP/IP reference model,
    • the ISO/OSI reference model, and
    • the hybrid reference model

“Philosopher-Translator-Secretary”-Architecture

Graphic by courtesy of Prof. Dr. Thomas C. Schmidt, HAW Hamburg

TCP/IP Reference Model or DoD Model

  • Developed from 1970 onwards by the Department of Defense (DoD) in the Arpanet project
  • Divides the required functionality to realize communication into 4 layers
  • For each layer, it is specified, what functionality it provides
    • These requirements are implemented by communication protocols
      • Concrete implementation is not specified and can be implemented in different ways
      • Therefore, for each of the 4 layers, multiple protocols exist
Number Layer Protocols (Examples)
4 Application Layer HTTP, FTP, SMTP, POP3, DNS, SSH, Telnet
3 Transport Layer TCP, UDP
2 Internet Layer IPv4, IPv6, IPX
1 Link Layer Ethernet, WLAN, ATM, FDDI, PPP, Token Ring

Described in RFC 1122 (TCP/IP)

TCP/IP Reference Model – Message Structure

  • Each layer adds additional information as header to the message
    • Some protocols (e.g., Ethernet) add in the link layer not only a header but also a trailer at the end of the message
    • The receiver analyzes the header (and trailer) on the same layer

Hybrid Reference Model

  • The TCP/IP reference model is often presented in the literature (e.g., by Andrew S. Tanenbaum) as a 5-layer model
    • Reason: It makes sense to split the Link Layer into 2 layers, because they have different tasks
  • This model is an extension of the TCP/IP model and is called hybrid reference model

OSI Reference Model

  • Some years after the TCP/IP reference model (1970s), the OSI (Open Systems Interconnection) reference model was developed from 1979 onwards
  • 1983: Standardized by the Intern. Organization for Standardization (ISO)
  • In contrast to the hybrid reference model, two additional layers are placed below the Application and above the Transport Layer

OSI Model Concepts

Central concepts of the OSI model are:

Services

Define what the layer does, i.e., its semantics

Interfaces

Define how to access it

Protocols

Describe how the layer is implemented

Physical Layer I

  • Transmits the ones and zeros

    • Physical connection to the network

    • Conversion of data into signals

  • Protocol and transmission medium specify among others:

    • How is the information encoded on the transmission medium?

    • Can transmission take place simultaneously in both directions?

Physical Layer II

  • At sender site: Signals are modulated onto the medium

  • At receiver site: Signals are demodulated from the medium

  • Devices: Repeater, Hub (Multiport Repeater)

Network Layer I

  • Forwards packets between logical networks (over physical networks)

    • For this internetworking, the network layer defines logical addresses (most commonly IP addresses)

    • Each IP packet is routed independently to its destination (\(\rightarrow\) connectionless)

Network Layer II

  • At sender site: Packs the segments of the Transport Layer in packets
  • At receiver site: Unpacks the packets in the frames from the Data Link Layer
  • Routers and Layer-3-Switches connect logical networks
  • Usually the connectionless Internet Protocol (IP) is used
    • Other protocols (e.g., IPX) have been replaced by IP

Transport Layer I

  • Transports segments between processes on different devices via so-called end-to-end protocols
  • Transport protocols implement different forms of communication
    • Connectionless communication, typically UDP (User Datagram Protocol) in TCP/IP networks
    • Connection-oriented communication, typically TCP (Transport Control Protocol) in TCP/IP networks

Transport Layer II

  • At sender site: Packs the data of the Application Layer into segments

  • At receiver site: Unpacks the segments inside the packets from the network layer

  • Addresses processes with port numbers

Combination of TCP/IP = de facto standard for computer networks

Session Layer

  • Controls the dialogues (connections) between processes
  • Provides the following services
    • checkpointing (and recovery)
    • authentication
    • authorization
  • Relevant protocols of the Session Layer are H.245, L2TP, PAP, and SOCKS
  • Session Layer services are commonly used for RPCs (cf. lecture Distributed Systems)

Many network applications do not require a dedicated session layer protocol.

Presentation Layer

  • Contains rules for setting the format (presentation) of messages
    • The sender can notify the receiver that a message has a specific format (e.g., ASCII) to make conversion happen, which is perhaps necessary
    • Data records can be specified here with fields (e.g., name, student ID number…)
    • Data types and their length can be defined here
    • Compression and encryption could be implemented by this layer

The functionality of the presentation layer is often implemented as part of the application layer.

Application Layer

  • Contains all protocols, that interact with the application programs (e.g., browser or email program)

  • Here is the actual payload (e.g., HTML pages or emails), formatted according to the used application protocol

  • Some Application Layer protocols: HTTP, FTP, SMTP, POP3, DNS, SSH, Telnet

wikipedia.org (CC0)

image

pixabay.com (CC0)

image

Reference Models – Summary

  • The OSI reference model is the most fine granular and is most widely used
  • Protocols of the physical and the data link layer are often highly entangled in practice
  • Many network applications do not require dedicated protocols on the session and presentation layer
    • Their functionality is often implemented as part of the transport or application layer

Topologies

Topologies of Computer Networks

  • The topology of a computer network…
    • determines how the communication partners are connected with each other
    • affects its reliability a lot
  • The structure of large-scale networks is often a combination of different topologies
  • Physical and logical topology may differ
    • Physical topology: Describes the wiring
    • Logical topology: Describes the flow of data between the terminal devices
  • Topologies are graphically represented with nodes and edges

Bus Network

  • All terminal devices are connected via a shared communication medium – the bus

  • No active components between the terminal devices and the shared communication cable
    • If a node fails, it does not affect the network itself
  • Advantage: Cheap to implement
    • In the past, Hubs and Switches have been expensive
  • Drawback: Shared communication cable fails
    \(\Longrightarrow\) Complete network fails
  • Only a single node can send data at each point in time
    \(\Longrightarrow\) otherwise, collisions will occur
    • A media access control method like CSMA/CD is required

Examples: (original) Ethernet, CAN, I²C, SPI

Ring Network

  • Connects node to node
  • All data is transferred from nodes to nodes until the destination is reached
  • Disruption of a single link \(\Longrightarrow\) network failure
  • Each node is also a repeater, which amplifies the signal
    • For that reason, large-sized rings (transmission medium dependent) are possible
    • Maximum ring length for Token Ring: 800 m

Examples:

  • Token Ring (logical): 4-16 Mbps
  • Fiber Distributed Data Interface (FDDI): 100-1000 Mbps

Star Network

  • All nodes are connected directly with a central component (Hub or Switch)
  • Failure of the central component leads to a failure of the network itself
    • The central component can be implemented in a redundant way
  • Failure of a node do not cause a failure of the network itself
  • Advantages: Expandability and stability

Examples:

  • (modern) Ethernet
  • Token Ring (physical): 4-16 Mbps
  • Fibre Channel (storage networks): 2-16 Gbps
  • InfiniBand (cluster): 10-40 Gbps

Mesh Network

  • Each node is connected with one or more other nodes
    • In a fully connected mesh network, the nodes are all connected to each other
  • If nodes or connections fail, communication inside the network is typically still possible because the frames are redirected
  • Advantages: Failure safe (depends on the degree)
  • Drawbacks: Cabling effort and energy consumption
  • Additional challenge: complexity to find the best way from sender to receiver (cf. Travelling salesman problem)

Examples:

  • Logical topology between Routers
  • Ad-hoc (wireless) networks

Tree Network

  • A dedicated root node exist with one or more edges
    • Every edge leads to a leaf node or to the root of another tree
  • Several star topology networks are hierarchically connected
  • Advantages:
    • Failure of a terminal device (leaf node) has no consequences
    • Good expandability and long distances are possible
    • Well suited for searching and sorting algorithms
  • Drawbacks:
    • When a node fails, the complete (sub-)tree behind is no longer accessible
    • In a large tree, the root may become a bottleneck because the communication from one half of the tree to the other half always needs to pass the root

Examples:

  • Connecting Hubs or Switches via an uplink port

Cellular Network

  • Implemented by wireless networks
  • Cell: Area where the nodes can communicate with the base station
  • Advantage: Failure of nodes do not affect the network itself
  • Drawback: Maximum dimension is limited by the number of base stations and their positions
  • Only one nodes can send data at each point in time \(\Longrightarrow\) otherwise, collisions will occur
    • A media access control method like CSMA/CA is required

Examples:

  • Wireless LAN = WiFi (IEEE 802.11)
  • Global System for Mobile Communications (GSM)

Current Situation

  • Today, Ethernet (1-10 Gbit/s) with Switches (\(\Longrightarrow\) star topology) is the standard for wired LAN
  • Connecting Hubs and Switches implements a tree topology, if there are no loops in the cabling
  • Cell topology is the standard for wireless networks
  • Mesh topology is one possible use case of wireless networks and it is the logical topology between routers
  • Bus and ring topologies are no longer used for new computer network infrastructures
    • 10BASE2 (Thin Ethernet) and 10BASE5 (Thick Ethernet) are outdated since the mid/end-1990s
    • May 2004: IBM sells his complete Token Ring product lineup

Summary

You should now be able to answer the following questions:

  • What is a Computer Network and what are its objectives?
  • What is the difference between bandwidth, throughput, and latency?
  • What is a reference model and what do their difference layers represent?