Fig 1.1 IEEE address format
These notes are divided into the following sections:
Introduction
Types of networks
Network addressing
Data communications
Segments, subnets and internets
Timing Relationships for Asynchronous,
Synchronous and Isochronous transmission
LAN management
Summary
Traditionally, Local Area Networks were just the link between PCs in a small organisation and were the transmission medium for relatively small amounts of data traffic e.g. email, ftp, client-server interactions
Increased local processing power has led users to demand more data per unit time from the network. Users are applying more sophisticated softwares that demand and can process more data than perhaps their current network can supply transparently, i.e. with little latency.
Multimedia applications are demanding ever-higher bandwidth with streamed audio and video, data and images.
This has led many existing legacy LANs to be overloaded as their capacity is sadly lacking. Those candidates due for upgrade or replacement are 10 Mbit/sec Ethernet which was been the installation choice for many LANs (due to among many factors its cheapness) and 4 or 16 Mbit/sec Token Ring which although was a more expensive networking solution was installed widely especially in the banking industry.
What industry wants are high bandwidth networks that provide a scalable network with a cost effective migration path i.e. such that the network is future proofed to some degree so that the network (or parts of it) may be upgraded even further with relative ease.
The ideal solution therefore is a scalable network where data rate, user number and size may be increased at some later date.
In order to design fast networks we must understand some of the operating
principles of today's fast networks.
The table below contrasts some characteristics of LANs, MANs and WANs
| LAN | MAN | WAN | |
| Size | Hundreds of metres to several kilometres, usually covering a building or group of close buildings | Tens of kilometres, covering a large town, city or perhaps military base | Hundreds to thousands of kilometres, covering a whole country, continent or the entire world |
| Data Rate | 0.2 to 1000 Gbit/second | 1.5 to 622 Mbit/second | 0.056 to 2488 Mbit/second |
| Terminals owned by | Users | Carrier or users | Carrier or users |
| Network equipment owned by | Users | Carrier or users | Carrier, e.g. BT, Mercury, Sprint etc |
| Cabling owned by | Users | Carrier or users | Carrier |
| Cabling type | Coax, twisted pair, optical fibre | Coax, optical fibre | Optical fibre, except for subscriber loop |
Table 1.1 Comparison of characteristics of network types
There are several differences between both the standards and types of networks in the USA and UK. Metropolitan Area Networks are more popular in the USA than the UK. In a WAN, it is likely that the cabling will be entirely Fibre Optic from end to end except for the subscriber loop (the section of cable from the subscriber's premises to the local exchange building) which is still twisted pairs of copper cabling. Many towns and cities are now cabled with FO to provide high bandwidth supplies to subscribers.
In any network, addresses are used to denote message sources and destinations otherwise there could be no communication beyond a broadcast message. Thus each end station needs a unique address. LANs can use locally or globally unique addresses whereas large internetworks and MANs and WANs require globally unique addresses. The addressing can be in two parts, a local part and a global part e.g. 35 High Street which will be the local unique part followed by some specification such as Wigan to ensure that the street address is unique. There are several globally unique addressing schemes that are managed by different international authorities e.g. IEEE, IETF, ISO, ITU-T (formerly CCITT) and the British Standards Institute. Within fast networks (the scope of this course) there are several different addressing schemes that may be used.
1 IEEE Addresses
IEEE addresses are most common in legacy LANs in which the hardware addresses are assigned to Network Interface Cards. NICs have a 6 byte address and this can be programmed using electrically re-programmable memory (EPROM). NICs are expansion cards containing the requisite electronics needed to interface the motherboard with the network. They are inserted into a free bus slot on the motherboard and provide the required software functions plus the hardware connections for the network. The IEEE addresses are 6 bytes long and include 2 flags of one bit each to denote whether it is an Individual or Group address (I/G) and whether it is a Universal or Local address. This is followed by the Organisationally Unique Identifier (OUI), issued by the IEEE and most often this will be the manufacturer of the network equipment e.g. IBM use 10005A and the ATM forum use 00A03A. There then follows a Serial Number (SN) assigned by the OUI owner to identify the individual machine within the LAN. Occasionally a manufacturer will supply two NICs with the same SN and this will cause problems when the network is initialised. Some NICs may allow the user to reassign the MAC number otherwise the card will have to be returned to the manufacturer for number re-allocation.
Fig 1.1 IEEE address format
2 Internet Protocol Addresses.
IPv4 is a common network protocol for large internetworks. IP addresses are OSI layer 3 (network) addresses and uniquely specify end points in large networks. They are 4 bytes long and consist of two fields to identify the network and the host. They are usually written in dotted decimal form e.g. 184.34.45.65. This gives a maximum possible number of 232 = 4 billion addresses but because of the demand for IP addresses there will be a shortage of unique identifiers in the near future, so to overcome this a new naming system that can address far more end points has been established. This is IPv6 and uses 128 bit (16 byte) addresses to provide a total of 3.4 x 1038 possible unique addresses which is (hopefully) sufficient for future predicted needs.
3 WAN addressing
The WAN community has developed several globally unique addressing schemes
and these will be useful in ATM LANs. ISO has several types of NSAPs (Network
Service Access Points). NSAP addresses are 20 bytes long and can encapsulate
other addressing schemes e.g. IEEE or IP.
Fig 1.2 NSAP addressing
4 ITU-T E.164 addressing
ITU-T E.164 addresses are essentially ISDN telephone
numbers and can be up to 15 digits long e.g. 01 – 304 – 732 – 1874.
|
|
Fig 1.3 ITU-T E.164 Addressing
The OSI model was originally published in 1979 with the objective of providing Open Systems Interconnection and eventually was adopted by some governments as the mandatory standard for government contracts. OSI defines 7 layers of functionality each with its own set of services and well-defined functions. This provides a common terminology and a framework for the development of standards. It enables manufacturers to provide different paths through the layers and enables interoperability of components from different manufacturers allowing customers a greater choice and lowering prices.
Work on TCP/IP had begun in 1974, so the OSI model was still a prototype
by the time that TCP/IP had already been proven useful and robust. OSI
can only stand as a yardstick by which we can compare other protocol stacks.
OSI detailed model
As shown in fig 1.4, the OSI model defines a path through the communication
structure of a computer system. The application a user is running e.g.
word processor will enter the model via the application layer if there
is a need to communicate with another remote machine.
PDU encapsulation
As data is transferred from layer to layer, each successive layer adds its own section of code to the data to identify it to peer layers. The layers are added to the protocol data unit until it reaches the physical layer. The raw bits are transmitted and at the receiving end, the extra data is stripped away from the PDU as it rises the protocol stack, finally leaving the data at the destination.
Figure 1.5 Protocol Encapsulation
Networks generally transport data in packets, frames or both. Packets belong to layer 3 and frames to layer 2. A packet generally has a header, payload but no trailer whereas a frame has header, payload and trailer
Messages or data that are too long to fit into a single frame will be segmented (broken into multiple units) for transport, see fig 1.6. This involves numbering each segment so that at the receiving end the segments can be reassembled to the correct order. Each segment will have its own header and trailer.
Depending on the mechanics of the delivery system, packets may need numbering to ensure correct order of reassembly upon reception of all the segments. Segmentation will also occur when a frame or packet has to travel across a network that has a smaller maximum size of frame than the source network. When the segments reach another network where the packet/frame size is larger, the reassembly process will recombine the segments into a larger frame.
Comparisons of Cells and Frames
ATM cells are of a fixed 53 byte size and are small when compared to a variable Ethernet frame (which can be 46 to 1500 bytes with a 14 byte header and a 4 byte trailer). The ATM cell is 48 bytes of payload with 5 bytes of header information. It has no trailer unlike an Ethernet frame. The cell size was deliberately chosen to be small. Datacomms wanted 64 bytes of payload whereas Telecom wanted 32 bytes so a compromise was reached with 48 bytes plus 10% for overhead i.e. 5 bytes. The reasoning for keeping the cell size small is to allow mixing of data and real time (isochronous) traffic without having to wait for the passage of a large Ethernet frame which may be in front. This may take 1.2 ms to pass across the network and would delay any real time applications running.
Figure 1.7 Comparison of ATM cells with Ethernet frames
Segments, Subnets and Internets
LAN segments are physical sections of a LAN that are connected by repeaters.
Many organisations have developed complex internetworks that combine several
different types of subnetworks. The subnets are usually groups of layer
3 (IP) addresses. Subnets are interconnected by an internetworking device
such as a repeater, bridge or router.
Fig 1.8 Diagram of LAN segment Interconnections
Internetworking
When connecting different networks there will have to be some consideration as to the characteristics of the networks in order to choose the correct internetworking device.
Fig 1.9 Differences between repeaters, bridges and
routers
Using Repeaters
Repeaters forward all frames regardless of the destination. This means that all segments will receive all frames regardless of whether the destination station is actually a physical part of that network segment. Repeaters re-time the signal.
Fig 1.10 Action of Repeater in Ethernet LAN segment
Many Ethernet LANs use multiport repeaters, otherwise known as hubs. Hub is a loose generic term and should be defined properly. Like the ordinary repeater, all frames are passed onto the next segment regardless of destination.
Bridges are used to interconnect similar LAN segments and can extend the size of a LAN beyond the limits imposed by repeaters. They also can isolate traffic to the segments in which it needs to be unlike repeaters. This can be useful when one part of a network is busier than the rest of the network e.g. design department passing large graphical files between each other. The traffic within the design department need not enter the rest of the network if it is addressed within the same department. Bridges selectively forward or filter frames based on OSI layer 2 addressing information. Bridges are network protocol independent. They do not examine the contents of the frames, merely the header information.
Fig 1.12 Illustrating the role of Bridges in traffic
filtering
Bridges are able to use spanning tree routing (IEEE 802.1d). This is the preferred method of routing in Ethernet. Bridges learn which stations are attached at which points by examining the source addresses on packets. Although multiple paths may exist, some bridge ports may be turned off to provide a tree without multiple paths. In times of breakdown, closed ports may be turned on again.
Another method here is source routing, preferred method in token ring, where stations use discovery frames to find routes to other stations. This is achieved by sending and receiving discovery frames. MAC addresses used here again.
Use of Routers
Routers are powerful devices that can interconnect arbitrary networks allowing any type of LANs to be connected, also allowing LAN to MAN and LAN to WAN connection. Routers operate at level 3, network layer and this requires high speed processing, complex software and extensive memory to cope with buffering. We now need IP addressing to send out of the network so stations need to be aware of informing routers of off-LAN traffic so that outgoing frames are sent to the router direct to relieve the network medium of excess traffic.
Fig 1.13 The role of Routers in network interconnection
Timing Relationships for Asynchronous, Synchronous and Isochronous transmission
For asynchronous traffic, source and receiver clocks will not be precisely related. Data is generally transmitted in bursts of frames. There is a preamble or start flag to allow the receiver to synchronise with the data rate of the incoming frame. This is common in LANs. Ethernet stations only synchronise when each frame is arriving as there can be unpredictable inter-frame arrival times.
Fig 1.14 Depiction of Asynchronous Transmission
Synchronous transmission precisely aligns the sender and receiver clocks. Data are usually packaged in repeating 125 ms frames, a common technique in packet switched networks.
Fig 1.15 Depiction of Synchronous Transmission
Isochronous traffic keeps a constant delay across the network. The source and receiver clocks are related and may be offset by a constant amount. Delay through the network should be constrained and constant delay is best, i.e. zero delay variation. Isochronous transmission is often used to deal with real time data such as voice, video, telemetry when an unexpected delay in delivery of data can result in loss of source intelligibility. This term is widely used when discussing ATM or voice/ video applications. Clocks are not permitted to vary in isochronous transmission.
WAN Frames
WAN transport schemes use identical time sized frames that recur at
125 ms with no interframe gaps. This provides
8000 frames per second. The benefits here are transport efficiency, scalability,
high data-transmission speed and both switching and mixing of asynchronous
and isochronous traffic. 8000 frames per second was chosen so that frequencies
of up to 4kHz may be digitised (Nyquist). The requirement was for 8 bits
in the frame giving a total of 64 kbit/sec for voice. WAN frames may have
more than 8 bits in each frame allowing scalability. See table 1.2 below.
| Data Rate Mbits/second | Bytes per Frame |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Table 1.2 Scalability of WAN data rates
According to the ISO, network management requires 5 functions.
SNMP is the de-facto standard for LAN management and is also supported by vendors of Fast-LAN equipment. It was designed by IETF to be simple. It uses IP and UDP packets. SNMP Version 1 is widely deployed and uses only 4 commands. Version 2 adds enhancements. It uses the client/ server model with agents and managers. Agents are client software and run in manageable devices such as stations, hubs, bridges and routers etc. Agents maintain standard Management Information Bases (MIB). Manager software runs on a management station and communicates with agents to both collect information and set values.
MIBs have a pre-defined data structure and send and store values such as the value of a port, on or off. TRAP can be used to send a message back to the manager that a user defined event has occurred such as network usage is over 40%.
Fig 1.16 Usage of SNMP
Explanations of Acronyms
Used
ASCII = American Standard Code
for Information
Interchange
ATM = Asynchronous Transfer Mode
CRC = Cyclic Redundancy Check
eprom = Electrically programmable
read-only
memory - a chip that can be easily reprogrammed
ICCTT= International Consultative
Committee for Telephones and Telegraphy
IEEE = Institute of Electrical
and
Electronic Engineers
IETF = Integrated Services Digital
Network
ISO = International Standards
Organisation
ITU-T = International Telecommunications
Union,
Telecommunications section
UDP = User Datagram Protocol