The Hermit Hermit's Free Library  Computer Networking
This article describes and compares popular networking architectures, including Ethernet, Token Ring, FDDI, ARCNet, and ATM.

Part 6: Network Architecture

(McFredries, p.990 / Andrews, p.901)

Each architecture is designed to overcome certain network problems, and each has its own advantages and disadvantages.

Different architectures use different approaches and all use different access methods to either prevent simultaneous access of the network cable, or to ensure error-free communications.

Comparison of Three Popular Network Technologies

  Ethernet Token Ring FDDI
Logical topology Bus Single ring Dual ring
Physical topology Star or Bus Ring or Star Ring
Media Twisted Pair, coax, or fiber-optic Twisted pair or fiber-optic Primarily fiber-optic
Standard Bandwidth 10 or 100 Mbps 4 or 16 Mbps 100 to 200 Mbps
How token is released N/A After receive After transmit
Max Nodes 500 260 1024
Advantages/Disadvantages Least expensive, simplest, most popular More reliable than Ethernet but more difficult to trouble-shoot Fastest of three. Reliable. Expensive, difficult to install and maintain


(Andrews, p. 903  / McFredries, p.990)

Ethernet is the most popular network technology today. Can be configured in either a Bus or a Star topology. Standard Ethernet throughput is 10Mbps but Fast Ethernet runs at 100Mbps.

Ethernet can use any of 6 cabling systems, of which the two most popular are 10BaseT in a star topology using a hub and 10Base2 (Thinnet) in a bus topology.

10BaseT can be used to connect two computers directly without a hub using a crossover cable . So-called because some of its connections are "crossed-over" to create the data path - something that is normally done inside the hub. I guess a cross-over wouldn't work satisfactorily to connect to a network via a hub.

Every Ethernet NIC has a unique address - its Media Access Control (MAC) address - burned into it by its manufacturer.

Ethernet is a baseband network. Baseband means that data is carried over the wire a single message at a time in digital form. Contrast baseband with broadband networks which send multiple (analog) messages over the wire simultaneously, each message on its own frequency.

Ethernet is also referred to as a passive network, which means that the networked computers and not some dedicated network device originate the signals used to manage network traffic.

Network traffic must be managed because if two or more nodes attempt to send data simultaneously a collision occurs and the potential exists for corrupted packets.

The method used by Ethernet networks to prevent concurrent transmission by more than one node is referred to as being contention-based because each node must contend for the opportunity to transmit on the network.

Specifically, the method used by Ethernet is CSMA/CD - Carrier Sense Multiple Access/Collision Detection. Translation:

It works this way:

Ethernet supports a relatively limited number of nodes before performance, speed, and reliability begin to drop. To overcome congestion, a network which has grown too large can be segmented using switches, bridges, or routers.

Nodes on a segmented network need only contend with transmissions within their segment OR transmissions which are destined for their segment. Local traffic within each segment is limited to that segment.

Token Ring

(McFredries, p.990  / Andrews, pp.907-9  / Keogh, pp.132-32)

Token Ring throughput is 16Mbps. Developed by IBM, Token Ring is more expensive and complex than Ethernet, but more robust and reliable. (Andrews, p.908).

Logically, Token Ring networks are rings, but physically nodes are connected in a Star topology. Nodes are connected to centralized devices (like a Star) which are patched together using their "Ring In" and Ring Out" connections, and form, along with the star-connections, the "ring".

There are several variations on the centralized device:

The first and last MAU in the circuit are connected by the Main Ring Cable, which may be fiber optic.

Each node contains a Token Ring LAN card which connects to the MAU using UTP or STP cables and a 9-pin connector. Each Token Ring NIC has a unique address, which is assigned during manufacture and burned into the card's firmware.

Collisions are avoided completely using the following method:

In addition to avoiding collisions, tokens increase reliability with the following method: (Keogh)


(Andrews, p.901)

Attached Resource Computer Network.

The oldest network around, it was developed in 1977 (Keogh). Seldom seen now according to Andrews, but it is still available.

Throughput of ARCNet was 2.5Mbps and it supported up to 32 nodes.

McFredries says it passes tokens from node to node using the numerical sequence of network addresses, but Keogh says that, like Ethernet, it uses Collision Detection Media Access Method (CSM/CA).


Fiber Distributed Data Interface ("fiddy"). (Andrews, p.909)

FIDDI runs at 100Mbps up to 62 miles (100km) and is often used as a backbone network. A backbone is a network which is used to connect several networks together. A single FDDI could, for example, be used to connect several Ethernet and Token Ring networks together.

FDDI uses either Fiber optic or, recently, UTP cable.

FDDI, like Token Ring is a ring-based network, but it does not use a centralized hub so it is both logically and physically a ring.

FDDI uses a token-passing method to control traffic, but it is more sophisticated and powerful method than that of Token Ring:

In addition, a FDDI network has dual, counter-rotating rings, unlike the single ring used by the Token Ring model. The purpose of the dual ring is to ensure that communications will continue even if a break occurs in the primary ring. It works like this:

Data normally travels on the primary ring, but if a break in that ring is encountered, the nearest FDDI station switches the data to the secondary ring and the direction changes so that data will return in the direction it came until it once again encounters the break. At that time it once again changes to the primary ring and switches direction again.

There is another option of FDDI networks called multiframe dialogues. Multiframe dialogues let one station send a limited token to one other station which sets up a private dialogue. A private dialogue permits continuous communications between the two nodes without interference from other nodes, which continue to pass the main token amongst themselves.

FDDI is expensive and difficult to install and maintain. (Keogh, p.135)


(Andrews, p.901,905  / Keogh, pp.167-71  / Sportack, pp.316-329)

Asynchronous Transfer Mode

Relatively new type of network - very fast and works well over short and long distances, although it is usually implemented on the WAN level.

Conversion of existing networks to ATM requires extensive (and prohibitively expensive) measures, therefore, ATM is usually implemented in one of two ways:

The big advantage of ATM is that it can provide different types of services over the same network:

Therefore, a company can use one network for data, telephone, and video conferencing instead of having to implement three separate networks.

ATM is a broadband network. Broadband networks send multiple messages over the wire simultaneously - each traveling on its own frequency in analog form.

On an ATM network the packets are referred to as cells. The cells are very small - only 53 bytes. The header is only 5 bytes of those 53, leaving 48 bytes for data.

The smaller cell size actually increases throughput because: