The basic reasons why computers are networked are

• to share resources (files, printers, modems, fax machines)
• to share application software (MS Office)
• increase productivity (make it easier to share data amongst users)

Take for example a typical office scenario where a number of users in a small business require access to common information. As long as all user computers are connected via a network, they can share their files, exchange mail, schedule meetings, send faxes and print documents all from any point of the network.

It would not be necessary for users to transfer files via electronic mail or floppy disk, rather, each user could access all the information they require, thus leading to less wasted time and hence greater productivity.

Imagine the benefits of a user being able to directly fax the Word document they are working on, rather than print it out, then feed it into the fax machine, dial the number etc.

Small networks are often called Local Area Networks [LAN]. A LAN is a network allowing easy access to other computers or peripherals. The typical characteristics of a LAN are,

• physically limited (< 2km)
• high bandwidth (> 1mbps)
• inexpensive cable media (coax or twisted pair)
• data and hardware sharing between users
• owned by the user

New cards are software configurable, using a software program to configure the resources used by the card. Other cards are PNP [plug and Play], which automatically configure their resources when installed in the computer, simplifying installation. With an operating system like Windows 95, auto-detection of new hardware makes network connections simple and quick.

On power-up, the computer detects the new network card, assigns the correct resources to it, and then installs the networking software required for connection to the network. All the user need do is assign the network details like computer name.

This allows intermediate devices in the network to decide in which direction the data should go, in order to transport the data to its correct destination.

A typical adapter card looks like,

A PCMCIA adapter card, suitable for connecting to a portable laptop computer to a network, looks like,

Peripheral cards associated with EISA and MCA are normally self configuring.

The major problem arises with cards for the ISA bus (found in the majority of AT type computers and clones). This is because the cards are configured by the user (using either jumpers or a software program).

Users make mistakes, and often configure cards so that they conflict with other cards already present in this system. This causes intermittent or immediate non-operation of the computer system.

For instance, a networking card that is allocated the same resources as a serial communications program may function perfectly, except when the user is logged into the network and then tries to use the serial port, at which time the machine will crash.

Resources Used By Peripheral Cards
We have already mentioned that resources used by ISA peripheral cards must not be shared (two cards cannot use the same). So what are the resources used by peripheral cards? Essentially, there are FOUR resources which are user configurable for peripheral cards. Some cards may only use one (a port location(s)), others may require all four.

The FOUR resources are

1. Input/Output Port Address
   In the PC, the port numbers used by peripheral cards range from 200h to 3FFh. The I/O port address is used by the PC to communicate with the peripheral card (issue commands, read responses, and perform data transfer).
2. Interrupt Request Line
   The interrupt request line is used by the card to signal the processor that the card requires the processors attention. ISA peripherals cannot share the same interrupt request line, and IRQ2 in AT/2386/486 computers should not be used (there are others which must also not be used). IRQ2 to IRQ15 appear on the ISA bus.
3. Direct Memory Request Line
   The DMA request line is used to transfer data between the peripheral card and the computers memory at high speed. DMA channel 0 cannot be used, as it is reserved for system use.
4. Buffer Memory Address
   Some peripheral cards prefer to use memory space rather than an I/O port address to transfer data to the processor. This memory space occupied by the peripheral card appears in the main system memory RAM area available to the processor (usually between C0000h to EFFFFh). Care must be taken to ensure this space is not being used for other purposes (like shadow RAM, EMS for windows, VBGA BIOS).

So How Do Peripheral Cards Work?
Peripheral cards require a software driver to function. This software driver provides the interface between the card and the operating system, making the services provided by the card available to the user.

The software driver is normally configured to match the resource settings of the card. This is done by a configuration utility, and stored either in the executable file, or a separate file (like .ini or .cfg).

It is obviously important for the configuration settings in the software driver to match those configured on the peripheral card.

The resources used by the card are either set by jumpers (or slide switches). New cards can also be configured using a software program, rather than by manually setting jumpers on the card. Where cards are software configurable, the cards retain their configuration when the power is turned off.

The software driver provides the follow functions

• initialization routine
• interrupt service routine
• procedures to transmit and receive data
• procedures for status, configuration and control

The basic operation looks something like,

• card receives data
• card generates interrupt by asserting interrupt request line
• processor responds to interrupt request and jumps to service routine
• service routine instructs processor to read data from port location
• interrupt service routine releases processor to continue previous work

The major problem is assigning values of these resources which are already being used by either the system or another peripheral card. It is therefore handy to know what the resources are which are used by common peripheral devices. The following tables identify these.

Common I/O Port Addresses

Common Interrupts

Common Memory Addresses

Common DMA Lines

Installing A Peripheral Card
This section discusses basic techniques for installing peripheral cards. By following standardized procedures, this will help to minimize damage to the system or peripheral card, and reduce the possibility of incorrect installation.

1. Determine the resources used by the computer
   Use the previous tables to determine the interrupts, memory and port addresses used by the current hardware in the computer.
2. Read the install manual
   Check the disk for a read.me file (and read it). Read the manual and take note of the jumper switches used by the card. Identify where these are located on the card.
3. Determine resources to be used by the card
   Allocate resources to the card which do not conflict with existing hardware.
4. Observe electrostatic protection in handling the card
   Use a wrist strap and ground yourself properly before handling the card. Handle the card by the edges. Do not touch the components or edge connector. Use electrostatic bags or an electrostatic mat.
5. Configure the card jumpers
   Set the jumpers on the card
6. Insert the card
   Remove the system base unit cover and insert the card into a spare peripheral bus slot. Observe electrostatic precautions.
7. Load the software driver
   If the card was provided with a software driver, install the software driver. This might involve running an INSTALL program, or copying the drivers to the hard disk. It might also mean adding the driver name to the config.sys file (DEVICE=xxxxx.sys).
8. Configure the driver software
   If the driver software needs to be configured (specify which resources the card is using), this information might be stored in a separate file (.ini or .cfg). Often, when installing the software, it will ask for configuration details. These must be the same as the hardware jumpers used by the card.
9. Test card (run diagnostics where provided)
   If the card was provided with diagnostic software, run that now to test the card and driver. This is a good way to test if the installation was done correctly.
10. Test the machine
    Test some of the other software packages on the system (like networking, serial communications and printing) to see if they still work. If they don't, this indicates a probable conflict of resources. In Windows 95 or NT, run the diagnostic program to check for interrupt and resource conflicts (MSD or WINMSD).

• ISA cards are a problem
• check what resources are already being used
• do not share resources between two cards
• interrupts can only be shared on EISA and MCA cards
• run the diagnostics software after installation
• if the computer hangs, remove one board at a time until the problem disappears

• define a generic voice and data wiring system that is multi-purpose and multi-vendor
• help minimize cost of administration
• simplify network maintenance and changes

A building wiring system covers a number of different elements

• horizontal wiring
• backbone wiring

Horizontal Wiring
The horizontal wiring extends from the wall outlet to the system center (telecommunications closet). It includes the

• the wall outlet
• the horizontal cable
• cables used to interconnect components [cross-connects or patch cables] in the telecommunications closet (TC)

Some general features of the horizontal wiring scheme are

Backbone Wiring The backbone wiring system interconnects telecommunication closets, equipment rooms and entrance facilities (i.e., the outside world). Some general features are

• star topology
• maximum of two hierarchical levels
• interconnections between any two TC must not go through more than 3 cross connects
• use of recognized media
• adherence to distance limitations

The latest version is 568B, which contains some enhancements to the original standard.

• 100-ohm unshielded twisted pair (not exceeding 800 meters)
• 150-ohm shielded twisted pair (not exceeding 700 meters)
• 50-ohm coaxial cable (not exceeding 500 meters)
• 62.5/125um multi-mode fiber (not exceeding 2,000 meters)

• four pair 100-ohm UTP
• two pair 150-ohm STP
• 50-ohm coaxial cable
• two-fiber 62.5/125um fiber

• pair 1: white-blue and blue
• pair 2: white-orange and orange
• pair 3: white-green and green
• pair 4: white-brown and brown

1. Work Area
   This is where the user is located. Patch cables connect the users equipment (such as phone, fax, computer) to the wall outlet.

• Category 1
  uses 22 or 24 AWG (American Wire Gauge Standard) solid wire and is not suitable for data transmission
• Category 2
  uses 22 or 24 AWG solid wire, used for PABX and alarm systems, and has a maximum bandwidth of 1Mhz
• Category 3
  uses 24 AWG solid wire having an impedance of 100 ohms with a maximum bandwidth of 16Mhz
• Category 4
  same as Category 3 but rated to 20Mhz
• Category 5
  uses 22 or 24 AWG pair wire having an impedance of 100 ohms with a maximum bandwidth of 100Mhz, typically using an RJ45 connector and used to implement 10BaseT

EIA/TIA-568A Connector Specifications

This is gradually being phased out in favour of EIA/TIA-568B

Recommended by the IEEE for 100Base-TX and T4 operation

• Desktop to Wall Outlet: 5 Meters
• Wall Outlet to Wiring Closet: 90 Meters
• Patch Cable in Wiring Closet: 5 Meters
• Fiber Optic Link between Wiring Closets: 2000 Meters
• UTP Backbone between Wiring Closets: 800 Meters

• Segment Lengths: Two pairs of UTP cable with a maximum length of 100 Meters.
• Cable Types: As the signalling frequency is 20Mhz, the minimum is Category 3 standard UTP.
• Connector Types: Category 5 type RJ45 connectors.

100Base-FX Wiring Standard Specifications
Fast Ethernet over Fiber Optic Cable

• Cable Types: Twin strands of Multimode fiber.
• Connector Types: SC, MIC and ST
• Segment Lengths: Full Duplex links are 2000 Meters, two switches or a switch adapter is 412 Meters. Repeater segment lengths can be a maximum of 320 Meters (typically less)

100Base-T4 Wiring Standard Specifications
Fast Ethernet for Category 3 UTP Cable

• Cable Types: Category 3 UTP
• Connector Types: RJ-45
• Segment Lengths: 100 Meters maximum.

• cost
• distance
• number of computers involved
• speed
• requirements [called bandwidth] i.e., how fast data is to be transferred

Shielded twisted pair uses a special braided wire which surrounds all the other wires, which helps to reduce unwanted interference.

The features of twisted pair cable are,

Unshielded Twisted Pair cable used in Category 5 looks like

Category 5 cable uses 8 wires. The various jack connectors used in the wiring closet look like,

The patch cord which connects the workstation to the wall jack looks like,

Distance limitations exist when cabling. For category 5 cabling at 100Mbps, the limitations effectively limit a workstation to wall outlet of 3 meters, and wall outlet to wiring closet of 90 meters.

All workstations are wired back to a central wiring closet, where they are then patched accordingly. Within an organization, the IT department either performs this work or sub-contracts it to a third party.

In 10BaseT, each PC is wired back to a central hub using its own cable. There are limits imposed on the length of drop cable from the PC network card to the wall outlet, the length of the horizontal wiring, and from the wall outlet to the wiring closet.

Patch Cables
Patch cables come in two varieties, straight through or reversed. One application of patch cables is for patching between modular patch panels in system centers. These are the straight through variety. Another application is to connect workstation equipment to the wall jack, and these could be either straight through or reversed depending upon the manufacturer. Reversed cables are normally used for voice systems.

How to determine the type of patch cable
Align the ends of the cable side by side so that the contacts are facing you, then compare the colors from left to right.

If the colors are in the same order on both plugs, the cable is straight through. If the colors appear in the reverse order, the cable is reversed.

The general features of coaxial cable are,

• medium capacity
• Ethernet systems (10Mbps)
• slighter dearer than UTP
• more difficult to terminate
• not as subject to interference as UTP
• care when bending and installing is needed
• 10Base2 uses RG-58AU (also called Thin-Net or Cheaper Net)
• 10Base5 uses a thicker solid core coaxial cable (also called Thick-Net)

Thin coaxial cable [RG-58AU rated at 50 ohms], as used in Ethernet LAN's, looks like

The connectors used in thin-net Ethernet LAN's are T connectors (used to join cables together and attach to workstations) and terminators (one at each end of the cable). The T-connectors and terminators look like

The features of fiber-optic cable systems are,

• expensive
• used for backbones [linking LAN’s together] or FDDI rings (100Mbps)
• high capacity [100Mbps]
• immune to electromagnetic interference
• low loss
• difficult to join
• connectors are expensive
• long distance

Fiber optic is often used to overcome distance limitations. It can be used to join two hubs together, which normally could not be connected due to distance limitations. In this instance, a UTP to Fiber transceiver [often referred to as a FOT] is necessary. Fiber optic cable looks like

In addition, fiber optic patch panels are used to interconnect fiber cables. These patch panels look like

When we think of how to send data from one computer to another, there are many different things involved. There are network adapters, voltages and signals on the cable, how the data is packaged, error control in case something goes wrong, and many other concerns. By dividing these into separate layers, it makes the task of writing software to perform this much easier.

In the Open Systems Interconnect model, which allows dissimilar computers to transfer data between themselves, there are SEVEN distinct layers.

Sending Data Via the OSI Model
Each layer acts as though it is communicating with its corresponding layer on the other end.

In reality, data is passed from one layer down to the next lower layer at the sending computer, till it's finally transmitted onto the network cable by the Physical Layer. As the data it passed down to a lower layer, it is encapsulated into a larger unit (in effect, each layer adds its own layer information to that which it receives from a higher layer). At the receiving end, the message is passed upwards to the desired layer, and as it passes upwards through each layer, the encapsulation information is stripped off .

• is a length of cable
• devices can be attached to the cable
• has its own unique address
• has a limit on its length and the number of devices which can be attached to it

Large networks are made by combining several individual network segments together, using appropriate devices like routers and/or bridges.

In the above diagram, a bridge is used to allow traffic from one network segment to the other. Each network segment is considered unique and has its own limits of distance and the number of connections possible.

When network segments are combined into a single large network, paths exist between the individual network segments. These paths are called routes, and devices like routers and bridges keep tables which define how to get to a particular computer on the network. When a packet arrives, the router/bridge will look at the destination address of the packet, and determine which network segment the packet is to be transmitted on in order to get to its destination.

In the above diagram, a packet arrives whose destination is segment B. The bridge forwards this incoming packet from segment A to the B segment.

Sometimes, a loop would be created which caused the wrong packets to be sent on incorrect segments. These packets could loop around the network, being forwarded on, eventually arriving back, only to be forwarded on, etc. This quickly floods the network. The spanning tree algorithm is a software algorithm which defines how switches and bridges can communicate and avoid network loops.

Packets are exchanged between bridges/switches, and they establish a single path for reaching any particular network segment. This is a continuous process, so that if a bridge/switch fails, the remaining devices can reconfigure the routing tables to allow each segment to be reached.

To be effective, ensure that the bridges/switches in use in your network support this protocol.

Repeaters also allow isolation of segments in the event of failures or fault conditions. Disconnecting one side of a repeater effectively isolates the associated segments from the network.

Using repeaters simply allows you to extend your network distance limitations. It does not give you any more bandwidth or allow you to transmit data faster.

It should be noted that in the above diagram, the network number assigned to the main network segment and the network number assigned to the other side of the repeater are the same. In addition, the traffic generated on one segment is propagated onto the other segment. This causes a rise in the total amount of traffic, so if the network segments are already heavily loaded, it's not a good idea to use a repeater.

Summary of Repeater features

• increase traffic on segments
• have distance limitations
• limitations on the number that can be used
• propagate errors in the network
• cannot be administered or controlled via remote access
• cannot loop back to itself (must be unique single paths)
• no traffic isolation or filtering

During initialization, the bridge learns about the network and the routes. Packets are passed onto other network segments based on the MAC layer. Each time the bridge is presented with a frame, the source address is stored. The bridge builds up a table which identifies the segment to which the device is located on. This internal table is then used to determine which segment incoming frames should be forwarded to. The size of this table is important, especially if the network has a large number of workstations/servers.

The advantages of bridges are

The disadvantages of bridges are

• the buffering of frames introduces network delays
• bridges may overload during periods of high traffic
• bridges which combine different MAC protocols require the frames to be modified before transmission onto the new segment. This causes delays
• in complex networks, data may be sent over redundant paths, and the shortest path is not always taken
• bridges pass on broadcasts, giving rise to broadcast storms on the network

Transparent bridges (also known as spanning tree, IEEE 802.1 D) make all routing decisions. The bridge is said to be transparent (invisible) to the workstations. The bridge will automatically initialize itself and configure its own routing information after it has been enabled.

Bridges are ideally used in environments where there a number of well defined workgroups, each operating more or less independent of each other, with occasional access to servers outside of their localized workgroup or network segment. Bridges do not offer performance improvements when used in diverse or scattered workgroups, where the majority of access occurs outside of the local segment.

The diagram below shows two separate network segments connected via a bridge. Note that each segment must have a unique network address number in order for the bridge to be able to forward packets from one segment to the other.

Ideally, if workstations on network segment A needed access to a server, the best place to locate that server is on the same segment as the workstations, as this minimizes traffic on the other segment, and avoids the delay incurred by the bridge.

Summary of Bridge features

Most routers can also perform bridging functions. A major feature of routers, because they can filter packets at a protocol level, is to act as a firewall. This is essentially a barrier, which prevents unwanted packets either entering or leaving designated areas of the network.

Typically, an organization which connects to the Internet will install a router as the main gateway link between their network and the outside world. By configuring the router with access lists (which define what protocols and what hosts have access) this enforces security by restricted (or allowing) access to either internal or external hosts.

For example, an internal WWW server can be allowed IP access from external networks, but other company servers which contain sensitive data can be protected, so that external hosts outside the company are prevented access (you could even deny internal workstations access if required).

Summary of Router features

• use dynamic routing
• operate at the protocol level
• remote administration and configuration via SNMP
• support complex networks
• the more filtering done, the lower the performance
• provides security
• segment networks logically
• broadcast storms can be isolated
• often provide bridge functions also
• more complex routing protocols used [such as RIP, IGRP, OSPF]

Nowadays, with the advent of 10BaseT, hub concentrators are being very popular. These are very sophisticated and offer significant features which make them radically different from the older hubs which were available during the 1980's.

These 10BaseT hubs provide each client with exclusive access to the full bandwidth, unlike bus networks where the bandwidth is shared. Each workstation plugs into a separate port, which runs at 10Mbps and is for the exclusive use of that workstation, thus there is no contention to worry about like in Ethernet.

These 10BaseT hubs also include buffering of packets and filtering, so that unwanted packets (or packets which contain errors) are discarded. SNMP management is also a common feature.

10BaseT Hubs dedicate the entire bandwidth to each port (workstation). The workstations attach to the hub using UTP. The hub provides a number of ports, which are logically combined using a single backplane, which often runs at a much higher data rate than that of the ports.

Ports can also be buffered, to allow packets to be held in case the hub or port is busy. And, because each workstation has it's own port, it does not contend with other workstations for access, having the entire bandwidth available for it's exclusive use.

The ports on a hub all appear as one Ethernet segment. In addition, hubs can be stacked or cascaded (using master/slave configurations) together, to add more ports per segment. As hubs do not count as repeaters, this is a better option for adding more workstations than the use of a repeater.

Hub options also include an SNMP (Simple Network Management Protocol) agent. This allows the use of network management software to remotely administer and configure the hub. Detailed statistics related to port usage and bandwidth are often available, allowing informed decisions to be made concerning the state of the network.

In summary, the advantages for these newer 10BaseT hubs are,

• each port has exclusive access to its bandwidth (no CSMA/CD)
• hubs may be cascaded to add additional ports
• SNMP managed hubs offer good management tools and statistics
• utilize existing cabling and other network components
• becoming a low cost solution

When a packet arrives, the header is checked to determine which segment the packet is destined for, and then its forwarded to that segment. If the packet is destined for the same segment that it arrives on, the packet is dropped and not retransmitted. This prevents the packet being "broadcasted" onto unnecessary segments, reducing the traffic.

Nodes which inter-communicate frequently should be placed on the same segment. Switches work at the MAC layer level.

Switches divide the network into smaller collision domains [a collison domain is a group of workstations that contend for the same bandwidth]. Each segment into the switch has its own collision domain (where the bandwidth is competed for by workstations in that segment). As packets arrive at the switch, it looks at the MAC address in the header, and decides which segment to forward the packet to. Higher protocols like IPX and TCP/IP are buried deep inside the packet, so are invisible to the switch. Once the destination segment has been determined, the packet is forwarded without delay.

Each segment attached to the switch is considered to be a separate collision domain. However, the segments are still part of the same broadcast domain [a broadcast domain is a group of workstations which share the same network subnet, in TCP/IP this is defined by the subnet mask]. Broadcast packets which originate on any segment will be forwarded to all other segments (unlike a router). On some switches, it is possible to disable this broadcast traffic.

Some vendors implement a broadcast throttle feature, whereby a limit is placed on the number of broadcasts forwarded by the switch over a certain time period. Once a threshold level has been reached, no additional broadcasts are forwarded till the time period has expired and a new time period begins.

• only the first few bytes of the packet are read to obtain the source and destination addresses
• the packets are then passed through to the destination segment without checking the rest of the packet for errors
• invalid packets can still be passed onto other segments
• there is little delay involved in packet throughput

Cut through switches use either a cross-bar or cell-backplane architecture.

• Cross-bar switches
• read the destination address then immediately forward
• acts as a simple repeater once the path is established
• can introduce delay: If the destination port is busy, it may need to buffer the packet
• Cell-backplane switches
• break the frame into small fixed cell lengths
• each cell is labeled with special headers which contain the addresses of the destination port
• the cells are buffered at the destination port
• the cells are then reassembled and transmitted
• the data rate on the backplane is significantly greater than the aggregate data rate of the ports
• in heavily overloaded networks, cell-backplane switching offers better performance than cross-bar switching

Store-Forward Switches

• examine the entire packet
• each incoming packet is buffered, then examined
• filters out any bad packets it detects
• good packets are forwarded to the correct segment
• detect more errors than the cut-through variety
• impose a small delay in packet throughput

Back Pressure Switches
Switches often employ buffering of packets. This is done so when packets arrive for a busy port, the packet is temporarily stored till the port becomes free. When the buffer becomes fill, packets become lost.

Back pressure switches overcome this problem by sending the overflow packets back to the workstation. This effectively slows the workstation transmission rate, and hence slows the arrival of new packets at the port.

Ethernet Switching: Advantges

• existing cabling structure and network adapters is preserved
• switches can be used to segment overloaded networks
• switches can be used to create server farms or implement backbones
• technology is proven, Ethernet is a widely used standard
• improved efficiency and faster performance due to low latency switching times
• each port does not contend with other ports, each having their own full bandwidth (there is no contention like there is on Ethernet)

Any traffic generated by these workstations can be sent to any other workstation in that domain. Workstations outside that domain are unable to see any packets (including broadcasts) that belong to the secure domain. Obviously, this has enormous implications for developing secure networks. Multiple virtual workgroups can exist, like email and www server. Users can belong to more than one virtual domain, thereby administration is centralized and security is maintained. The use of switch technology makes this possible.