This guide will cover the following areas.

Introduction: The two main types of network

A computer network is the linking of two or more devices for the purpose if effective communication.  There are two main types of computer network - both have a number of elements in common.  Each device on the network is considered a node and these nodes are connected in some way to enable communication.

The first type of network that a new user is likely to encounter is called a peer-to-peer network.  In this type, no node is considered more important than another is and each one has equal access to network resources.  There is no single node that is able to control the network.

Examples of peer-to-peer networks include:

  • PC to Printer
  • PC to PC, (i.e. Internet share cross-over link)

A peer-to-peer network is useful only up to a point.  After about the twelve-user mark the network will start to experience increasing difficulties in administration and control.  However, for low traffic file sharing on small networks this type can be ideal.

The second type of network that the user is likely to encounter is a Client-Server network.  With this type, the network is controlled by one or more servers with all other nodes acting as clients.

This networking type has distinct advantages allowing the network to become much larger while adding to the functional benefit of a network. 

Some examples of Client-Server networks include:

The primary advantage of Client-Server is this: control

Client-Server allows control over users access to network facilities and data.  This effectively means that networks, with the added security of control, are now a viable, large-scale business tool.

 

Network Topologies

Any given network can be said to have a topology of one sort or another.  A network’s topology is the logical layout of the connectivity of the nodes.  The exact configuration of the topology will affect the type of protocol or rules used in network communication.

For the purpose of this guide, we will consider that there are five main types of network topology.

Bus

With a bus-topology, a single main line or backbone runs the length of the network.  Each node connects to this backbone and all communication must happen over it.  This means that only one device can transmit at any given moment.  When more than one device transmits at the same time the message becomes garbled and a data collision is said to occur.

 [PC] [PC] [PC] [PC]
  |    |    |    |
|-+----+----+----+-|



At each end of the bus is a terminator that will indicate the end of the line stopping messages from endlessly travelling up and down the bus.

Having only one device able to transmit at a time clearly limits the scalability of bus networks but this has not stopped bus style architecture from enjoying a time of unrivalled success as the cheapest and most efficient network architecture.  It was here that Ethernet standards developed. 

Due to the drawbacks of bus architecture as an organisation grows in size they must employ ever more ingenious methods of maintaining the network or move to another architecture.

Ring

The ring-topology answers some of bus architectures biggest drawbacks by adding a method of control.

Here is a simple ring:

[PC]----------[PC]
 |             |
[PC]          [PC]
  \____[PC]____/

Under the token-ring protocol, (the main implementation of early rings), a token is passed from computer to computer giving each the chance to transmit.  This removes the data collision, but for larger networks the data rate slowly drops with each additional node.

Ring topology also suffers from another physical limitation – just as a bus requires two terminators, the ring requires an unbroken loop or effective communication is broken.

Mesh

The mesh-topology provides durability lacking in ring and scalability lacking in bus.  Under a mesh-topology, each node is connected to many other nodes.  The main communication device must be a level three device capable of making complex routing decisions for the delivery of data packets.

The main draw back is the financial practicality of implementing a mesh network.  There is no centre of control and the network may become fragmented and repetitious with infinite-loops, dead-ends and lost packets clogging up the system.

Mesh-topology works best with MAN and WAN sized networks like JANET, Kentish-MAN and the Internet.

  • MAN: Metropolitan Area Network

  • WAN: Wide Area Network

  • JANET: Joint Academic NETwork (a UK MAN formed between leading universities)

  • Kentish-MAN: A MAN project in Kent, England

  • Internet: A network of networks that allows such things as: the World Wide Web and universal email.

Star

Star-topologies solve the complexity of mesh while maintaining some of its better features. 

In a star-topology all of the nodes are connected to some form of central hub this is usually a switch or sometimes a router, but in times not so long passed it may have been a multiport repeater.

        [PC]
         |
[PC]     |       
  \____[hub]___[PC]
       /   \
      /     \
    [PC]   [PC]

A star is limited only by the number of ports in the hub device and the quality and aforethought of the rooms cabling.  Stars are perfect for networks that fit into a room or two.

The moderately gifted reader will recall that it has been suggested that the better features of mesh are retained by star and this is true for a hierarchy of computer systems can be created where are star network ends not in a PC or similar node but with another star network.  In this way, a pyramidal hierarchy that sometimes closely mimics management structure can be achieved.

Hybrid or Tree

However, when the powerful simplicity of star is combined (with modern communications equipment) with the ideas of each of the other topologies is the true power of networking realised. 

This method is sometimes called tree as the diagram of bus plus star looks very much like a tree.

In this illustration we can see five different servers making use of a bus topology (some via a ring) in order to serve users on a star network.

            [Accounts]---[CAD] 
[Users]           |        |
   V              |        |
  >+<          [File]-+-[Mail]
   |     [WWW]        |
   \________|_________/

The device that makes this possible is called a bridge.  A bridge device (including many routers) links two networks and translates between protocols.

In this situation it is important to realise that the bus would not be the now redundant 10 Mega-bite Ethernet but is likely to be fibre and running at a gigabit or more.  Similarly, the ring for the servers is likely to be notably faster than the upper limit required by the network.

 

Network Operating Systems

  • What is a network operating system?
  • Different kinds of NOS and their respective traits.

What is a network operating system?

A network operating system (NOS) or “extended operating system” is much like any other with respect to the workstation but it has the added power to extend the influence of the computer onto the network.  A good NOS will cause the computer to become less like an independent device and more a part of a great whole.

Different kinds of NOS and their respective traits.

There are a large number of network operating systems available for the average computer. 

  • The Windows Suite
  • Linux, Unix and the like
  • Others including Mac

The Windows Suite

This include NT, 2000 and the many different server specific OS.  The windows suit of server applications and specialised server operating systems present a dazzling array of choices to the inexperienced network manager.  Most have some use, many are best forgotten and some are actually worthwhile.

With so much potential power and so many specialised server OS in the Windows range it is unsurprising that Windows is considered very flexible and horribly vulnerable.  The situation is not as bad as it once was but huge gaps are discovered daily meaning that implementation is usually steady, deliberate, carefully considered and usually really quite simple. 

Compared to the leading rival (Linux) windows is by far the easiest to configure and clearly in second place security wise.  However, despite some shortcomings Windows is well documented and supported.  Windows also benefits from being familiar to many users.

Linux, Unix and the like

Often simply called “*nix” this range of OS are developed free of charge under the banner of open source development.  From the onset this range of OS were designed to be multi-user, network operating systems.

Linux now comes in a number different “flavours” some of which are better suited to different tasks. Debian is generally considered to make the best server. There is no range of OS just one all purpose freely available and highly acclaimed OS. Debian Linux is considered to be the most stable and secure of the Linux distributions. However many people find the slower and more through pace of development not to their taste.

Others including Mac

There are other less well-known NOS including Novell Netware and Mac OS for the Apple Mac range of PCs. (There is also a Linux designed for Macs).

There will always be marginalized alternatives some of which may gain popularity (or at least refuse to die like Novell netware) while others (like a certain abortive Apple-Linux) fall by the way side. Almost all make use of the same protocols, (Namely TCP/IP), and are designed to use hybrid and star topologies and do so more effectively than on others.

 

Conclusion: Summary of networking

It can be said, therefore, that there are three main areas of consideration with any network:

The choice in any category will influence heavily the available choices that can be made in the others.

 

Further Reading:

The budding network designer may wish to additional review: