Most of you have some sort of networking in place in your department. The networking in nuclear medicine, radiology and radiotherapy departments are usually done in a piece meal fashion. Everytime a new modality arrives it comes with a hub or a switch that needs connectivity to the rest of network. The result is the mess below!
A crazy quilts of hubs, switches and router spanning many buildings within and sometime outside the nuclear medicine department. This was our network over a year ago spanning from radiology to radiotherapy department. This network has no structure. By structure I mean the network has no underlying order that make it easy to understand , visualize and manage. The basic rules of the employed technology have been broken. It was slow and errors in transmission a common occurrence.   Eventually a major redesign and restructuring will become necessary.   One of the first thing you need to consider is which network technology to go for. There are four main network technologies that you could choose from.
The last two are dead. ATM is superior, inappropriate, expensive dying . Ethernet remains the dominant technology and the most appropriate for Nuclear medicine.

There are four variants to Ethernet technology:
1.Standard Ethernet operating at 10 Mbps. Depending on type of cables used are referred to as 10Baset, 10Base2 or 10Base5
2. Fast Ethernet operating at 100 Mbps.
3. Gigabit Ethernet operating at 1000 Mbps
4. And finally 10 Gigabit standard , due for ratification this month

Ethernet uses a communication method known as Carrier Sense Multiple Access/Collision Detection (CSMA/CD). Information is transmitted in Frames or packets of data. All devices share the same channel of communication and only one device can have a frame on the network. Each device listen to the network for collision before transmitting a frame or packet of data. If two workstations transmit at the same time, then a collision will occur. Each workstation will then wait for a random amount of time and try again. As the number of workstations increases, collisions increase near to an expotential manner, a real carnage of frames. Ethernet is a suicidal technology. Reducing this unnecessary carnage is one the most important criteria in designing an efficient and well tuned network.   In order to do this we need to know the types of frame that wiz around an Ethernet network.
There are 3 types of frames:
1. Unicast Frames. These are ordinary data frames that are addressed to a specific device or workstation .
2. Broadcast Frames which are broadcasted to all devices on the network. All devices broadcast their existence and services bys sending broadcast packets of data.
3. Multicast frames are addressed to a group of workstations
    There are three basic network devices that handle these frames on their way to their final destination. These devices are known as network building blocks
Repeater is a device that enhances the signals and repeat them. It lets all packets through. In another word it is frame stupid. It is used to extend the allowable length of cable.   Bridges have some sort of intelligence in that they are aware of frames as pockets of data and could detect the Ethernet or MAC address of both sender and recipient devices. They are not frame stupid but they are protocol stupid. They do not understand protocols such TCP/IP and IPX/SPX. They do not know how to handle broadcast framse because they do not carry the recipient Mac or Ethernet address.
Routers are intelligent devices that can inspect a packet of data and make a decision how to route to other networks.
  These components although originally existed as physical units, it is now becoming more and more as logical devices. The functionality of network devices is described on the basis of where they operate in OSI model.
  For two computer to communicate they both must conform to a common set of complex rules and regulation known as protocols.
The International Standards Organization (ISO) breaks down this complexity into 7 layers which is used as a template for creating interoperable hardware and software.
Repeaters operate at layer 1 – physical layer
Bridges operate at layer 2 – Datalink layer
Routers operate at layer 3 – Network Layer  
All Network devices in the market are described by the OSI model …. Next slide
1. Hubs which are multi-port repeater are described as layer 1 devices – they let all pockets to go through.
2. Switches which are basically multi-port bridges are describes as layer 2 devices. They could be used to divide the network into separate collision domains, but they can not stop broadcast frame. But there are switches that can operate at layer 3 and can stop and confine the broadcast frames to a workgroup a functionality known as Virtual LAN or VLAN
3. Multiport switched routers are described as layer 3 devices. These are silicon-based routers as against conventional processor based routers. Because they are silicon based they could operate at wire-speed that is 1000 of times faster that conventional processor based routers. (they use ASIC or Application Specific Integrated Circuit)
4. There are multi-port switched router that operate at level four, they could filter packets of data based on the application type.
A large Ethernet network is shaped by combing these components. There are rules as to how we employ theses devices.
Each variant of the Ethernet technology has its own set of rules which will determine how you deploy routers , switches and bridges. There are rules on maximum:
1.Length of cables
2.number of workstation per segment
3.Number of repeaters, segments in any path between two stations
4.Number of populated segments
  For example for standard 10 Mbps Ethernet there is a rule known as 5-4-3 rule which says:
the path between any two computers within a logical network may not include more that
• 5 segments and /or
•4 repeaters and/or
•3 segments with computer on them
As we move from 10 Mbps to 100 Mps the rule changes and become more strict, 5-4-3 rule no longer applies and loses its meaning. 5-4-3 rule is replaced by network diameter rule which applies to all variants of Ethernet technology.
The network diameter is defined as the wire distance between two end stations. For a standard 10BaseT network using twisted pair cables, the maximum allowable diameter is 500 m. As we move to Fast Ethernet the maximum allowable diameter becomes just 205 meter.
Both these networks have one collision domain. Performance drops drastically as you increase the number of hosts . You need to divide the network into separate collision domains.
This is a standard 10baseT network with maximum diameter of 500 meter but has one collision domain. We could divide it into 4 separate collision domains by collapsing the backbone using a switch or multi-port bridges.   This reduces the network diameter from 500m to 200 mm We could increase the network diameter by having a backbone consist of many switches.   The efficiency of this network could be further increased by moving our server to the higher level of hierarchy and connecting it direct to the backbone switch, giving them their own collision domain.