B227 Lecture Overheads - Topic 4: IEEE Standard 802 for LANs

IEEE Standard 802 for LANs

A series of standards by IEEE in very wide-spread use for LANs
We will look at three of the standards

802.3 CSMA/CD,

802.4 Token Bus, and

802.5 Token Ring

IEEE 802.3

Commonly referred to as Ethernet.

The name Ethernet comes from "luminiferous ether", which physicists used to think is a substance by electromagnetic radiation is transmitted through. Ethernet was actually a standard by Xerox, DEC and Intel, which 802.3 was based on. 802.3 actually contains a whole lot more than what the original Ethernet specifications had. These days, the term Ethernet usually refers to IEEE 802.3.

The original standard defines a 10Mps. Today, Ethernets can support 1Mbps or 100Mps.

A 1-persistent CSMA/CD LAN
802.3 defines 2 categories: baseband and broadband

Base band refers to digital signals, and broadband refers to analog signals

There are currently a few different baseband LAN categories:

10Base5 (also called Thick Ethernet, or Thicknet)
10Base2 (also called Thin Ethernet, or Thinnet)
10Base-T
10Base-F
100Base-T and 100Base-F (known as Fast Ethernet - more on this in the lecture on High-Speed LANs)

We also use the category names for the special cables they use (eg. 10Base-T cables). The first number refers to the data rate in Mbps. The last character (5, 2, 1, T or F) either indicates the maximum length of the cables in hundreds of metres (so 500, 200 or 100 meters – using coaxial cables) or the type of cable, twisted pair or fiber optics.

Figure 4-18 textbook p278

With 10Base-T and 100Base-T, we wire the stations to a hubs - the function of passive hubs is to receive a signal from any of the incoming lines, then amplify and broadcast out to all other lines (probably one of them will be connected to the server). We also have switching hubs, which actively redirect the signal traffic at appropriate speeds (eg 10Mbps or 100Mbps).

With 10Base5, we have transceiver connections on each station – transceivers does the carrier and collision detection, as oppose to the hubs

With 10Base2, we have passive BNC T-connectors on each station. The transceiver electronic are on the card in the station.

With most Network Interface Cards (NIC) these days, you will see two types of plugs – one is the BNC T-junction which is like a cylinder, and the other is for RJ45 connectors which looks like telephone jacks. The RJ45 connectors are for twisted pair wires to hubs.

To form networks larger than the specified cable lengths (500, 200 or 100 meters), repeaters are used.

Manchester Encoding

802.3 doesn't use normal binary representation for signaling. the problem is that if we have eg. 0 volts for bit 0, and 1 volt for bit 1, we won't be able to tell when a line is idle (0 volts) and when it is transmitting bit 0 (also 0 volts). Having two non-zero voltages for 0 and 1 (eg 1 volt for bit 0, and 2 volts for bit 1) is not a solution either since we end up with a continuously high DC (direct current) voltage value.

Digital signals in 802.3 use Manchester or Differential Manchester encoding.

Figure 4-20 p280

In Manchester Encoding, we represent 0 as a change from a low voltage (-0.85 volts) to a high voltage (+0.85 volts) during a certain time interval. We represent 1 as a change from a high voltage (+0.85 volts) to a low voltage (-0.85 volts). Both changes results in a DC value of 0 volts, which is ideal.

Note that this is a physical layer specification. It deals with using the voltage of a line to represent 0's and 1's. IEEE 802 specifications include physical layer issues as well as DLL services.

The 802.3 Frame Format

Figure 4-21 p281

Special multicast or broadcast addresses for destinations

Frames with the broadcast address goes to all stations on the LAN. Frames with multicast addresses goes to only a group of stations in the LAN. One station can belong to different multicast groups.

All frames have a minimum length 64 bytes – shorter frames are padded.

This is to prevent the whole frame from being sent out before the station manages to detect a collision.

The 802.3 Back-off Algorithm

Stations use the binary exponential back-off algorithm if it detects a collision when transmitting a frame.

If station detects a collision, it selects randomly whether to back off 0 (ie, retransmit immediately) or 1 time slot.

On the second time, if it again detects a collision, it selects randomly to back off 0, 1, 2, or 3 time slots.

On the third time, if it again detects a collision, it selects randomly to back off 0, 1, 2, 3, 4 5, 6, or 7 time slots.

On the n^th iteration, it selects between 0-2ⁿ -1 time slots to back off.

This keeps goes, until the maximum of reaches 1023 time slots. Then it keeps selecting between 0-1023 time slots to back off every time there is a collision.

Although extremely unlikely, it is possible that a station can wait forever to send a frame.

The 802.3 Performance

Figure 4-23 p285

Switched 802.3 LANs

A switch LAN allows us to mix different speed network cards in a single LAN.

Figure 4-24 p286 shows an example switch.

In this example, we have a switch with 4 plug-in cards- the vertical slots. Each plug-in card contains connectors to one particular type of NIC (eg. one plug-in card might be for 10Base-T, another plug-in card for 100Base-T). So different stations with different NICs will connect up to the appropriate plug-in card on this switch.

When a switch gets a frame from one particular connector on one particular plug-in card. It determines whether the frame is meant for another connector on the same card. If it is, it sends it out on that connector. If not, it sends (switches) it to another plug-in card through a high-speed backplane (not shown in the figure - the backplane connects all the plug-in cards in the switch).

In normal 802.3 set-up, all stations contend during a certain contention period - we refer to this as having one collision domain. In a switch, we can restrict the contention to only stations connected to one plug-in card (in which case we say that each plug-in card has its own collision domain), or the plug-in card can buffer up all incoming frames to transmit later (in which case EACH STATION has it's own collision domain, which means NO COLLISIONS HAPPEN).

IEEE 802.4: Token Bus

Drawbacks of Ethernet (802.3):

Worst-case delay is unbounded
Although highly unlikely, it is possible based on the binary exponential back-off algorithm that a station will wait infinitely long to send a frame

No mechanism for priority
We can't specify which frames are more important, and should therefore be transmitted first when a channel is seized.

Very important for time-critical frames like those containing voice or video.

Solution: Token Bus

Figure 4-26 Tanenbaum textbook p289

IEEE 802.4: Token Bus

The stations can logically be considered to be in a ring, although physically are connected as a bus network.
How it works:

Initialize network
First station sends first frame, then sends a token frame to next station in the ring
Only the token holder can transmit - no collisions
Token holder passes the token when finished
The order of the stations in the ring is irrelevant

Since the network is a broadcast network, every station receive all frames, including the token frames. So the token passing doesn't have to follow the physical layout of the stations. Eg. in figure 4-25, we can have an arbitrary logical ring order of 11, 20, 13, 17, 7.

802.4 Frame Format

Start and end bytes for framing - no need for length field
For frames containing data:

Frame control field will have values, 0, 2, 4 or 6 - for priority
When a station gets a token, frames of higher priorities (6 being highest) are transmitted first

Each station has a fixed time to transmit all frames - this ensures that there is a limit to the worst-delay for all stations.

If not all the frames of a station are transmitted during that fixed time, it will have to leave it for the next time the token comes around. The use of priority field ensures that the more important frames are guaranteed to be sent first.

For control frames:

Frame control field will have one of the 7 values

See Figure 4-27 p290.

These control frames are sent for maintaining the ring in situations of:

Passing a token
Establishing the ring
A station entering the ring
A station leaving the ring
Re-linking the ring when a station goes down

IEEE 802.5: Token Ring

Advantages of rings:
A collection of point-to-point links - well-understood and field tested technology
Rings do not need analog components (eg. voltage testing for collision detection in 802.3)
The use of the channel is fair for all station

Ring Networks and Interfaces

Figure 4-28 p293

At ring interfaces:

Bits copied in onto a 1-bit buffer, then copied out again
The stations can process this buffer

Like Token Bus, the station must seize the token to transmit

This is done by inverting one bit in the token before copying it out

Note here that unlike the analog carrier sense protocols we discussed in CSMA/CD and 802.3 and 802.4 (where every station detects the bits AT THE SAME TIME), in Token Ring the bits are passed from one ring interface to the next. Any particular bit would be only located at one particular station at any single time.

The ring must have sufficient delay to contain a token

The first bit of a token must not have come back to the same interface before it transmit out the last bit.

The delay is based on the interfaces' 1-bit delay, and the signal propagation time

If this delay is not sufficient, artificial delays must be introduced.

Must also consider when the network starts to lose machines (eg. workstations with interfaces being switched off at the end of the day), there will be less 1-bit delays.

A Wire Center in a Ring Network

Figure 4-29 p295

Star-shaped rings
Having bypass relays at a center to complete ring circuit if a station goes down
Relays triggered by lack of current from a station
Improves reliability and maintainability

Can use relays to test the performance of individual stations, to find faulty stations or line segments.

Token Operations of 802.5

When no traffic, a 3-byte token (Figure 4-30a) circulates endlessly
When a station seizes the token,

It converts 1-bit in the 3rd byte of the token, changing it to the first three bytes of a data frame (first 3 bytes in Figure 4-30b)
It then transmits the rest of the data frame
When finished, it transmit the token frame again.

Ring Maintenance of 802.5

One stations is always elected as a monitor.
Function of the monitor station:

Ensure token is not lost
Taking action when ring breaks
Clearing the ring when frames are garbled
Clearing orphan frames - orphan frames are first parts of frames generated by a station which crashes in the middle of its transmission.

Inserting delays where necessary - so a 3-byte token can circulate in the ring

If a monitor goes down, all other stations goes through a contention algorithm to pick a new monitor.

This philosophy of having a central station, which so much of the network depends on, is completely different to 802.4. The 802.4 guards very heavily against crashes, and so will not depend on a particular station. This approach is mainly due to the fact that 802.4 was designed for computers controlling factory production lines, where a crash can incur very heavy damages and costs.

Comparison of IEEE 802 LAN standards

802.3 Ethernet:

Advantages:

A lot of operational experience with it - very widely used

Simple protocol
Stations installed on the fly
Delay at low load can be zero - no need to wait for token

Disadvantages:

Requires analog circuitry - for collision detection

Minimum 64-byte for all frames - a lot of overhead for very small data packets

Worst-case delay unbounded
No priorities for frames
At high loads, collisions cause a major drop in throughput

802.4 Token Bus:

Advantages:

Uses very reliable and easy to get cable TV technology
Worst-case delays can be better determined - although it can be pretty bad in cases where the token gets lost repeatedly
Low overhead on short frames
Supports priority for frames
Excellent throughput at high loads

Disadvantages:

Uses a lot of analog components (modems and amplifiers)
Very complex protocol
Small base of users
Must wait for token, even at low loads

802.5 Token Ring:

Advantages:

Easy engineering
Fully digital
Use of wire centers makes it the only LAN that can detect and eliminate cable failures automatically
Low overhead on short frames
Supports priority for frames
Excellent throughput at high loads

Disadvantages:

Use of monitor stations - critical that the monitor remains operational
Must wait for token, even at low loads

IEEE 802.2: Logical Link Control

802.3, 802.4 and 802.5 doesn't worry about error control - it depends on higher layers to do it
IEEE defined 802.2 to run above it LAN protocols, to provide a consistent interface to the network layer

Having this LLC means that the protocols on the network layer can use any of the 802 protocols in the same way.

Figure 4-33 p304

[Back to Lecture Notes page]