Personal tools
You are here: Home Classes Fall 2004 - Spring 2005 CS 342 LAN standards; Ethernet
Navigation
Log in


Forgot your password?
« July 2008 »
Su Mo Tu We Th Fr Sa
12345
6789101112
13141516171819
20212223242526
2728293031
 
Document Actions

LAN standards; Ethernet

by admin last modified 2005-05-12 17:54

Wired LANs

The IEEE 802 Standard is a standard for local area networks.
  • 802.3  Ethernet
  • 802.4  Token Bus
  • 802.5  Token Ring
  • 802.11  Wireless
  • 802.15  Bluetooth
  • 802.16  Broadband wireless

Standards are concerned with
  • Addressing
  • Frame structure
  • Transmission and coding
  • Cabling/topology  (physical aspects)

Today we will focus on various aspects of Ethernet and the IBM Token Ring networks.

Ethernet

Hardware technologies

10 Base 5    thick coax     (the original)
10 Base 2     thin coax   -- Use T connector at each adapter to tap into signal, terminator at each end.
10 Base T     twisted pair  --  Use hub, individual connections (a star topology at the physical level)
10 Base F     fiber

The "10" indicates 10 Mbps transmission rate
"5" indicates 500 m(eters) the maximum distance between nodes without repeaters.
Base = Baseband signalling

10 Base 2

Can use up to 4 repeaters, extending length to 985 meters

10 Base T  --  Limited to 100 m with twisted pair

100 Base T -- Uses higher quality shielded twisted pair for higher data rate

10 Base FP  (fiber)

100 Base F(P, G, or B)

Gigabit Ethernet

Encoding

Digital transmission  ("baseband signalling"), so one channel uses all bandwidth.

Manchester encoding:



1 is encoded as high-low

0 is encoded as low-high

This allows a receiver to resynchronize to a sender.  But it uses twice as much bandwidth as simple binary would.


Addressing

Because messages are broadcast, each station needs an address (to identify the intended recipient).

Unique 48-bit address burned into the Ethernet adapter's ROM.  Address space managed by IEEE.

The first 24 bits are a block number.

A manufacturer can request a block of addresses of size 224 for a nominal fee.

Address does not change if a PC is moved from one LAN to another.

An address of 1111......1 (all 1's) indicates a broadcast message.

ARP (Address Resolution Protocol) is used to map IP addresses to Ethernet addresses.


Ethernet frame structure




1.  Preamble (8 bytes)

In Ethernet, all 8 bytes look like 10101010.

In 802.3, the first 7 bytes look like 10101010, the 8th like 10101011.

This marks the start of the frame, allows receiver to get in synch.

2.  Destination address (6 bytes)

3.  Source address (6 bytes)

4.  Type (2 bytes)

This is a code to indicate which network protocol is used in the data field (IP, IPX, etc.) so receiving Ethernet layer knows which module to pass it up to.

Supports the idea that one host can support more than one network protocol.

5.  Data field (46-1500 bytes)

Too short a frame may allow a sender to complete a transmission before a colliding packet arrives.  In that case, the sender would not know that there was a collision and would not resend the frame.

6.  CRC (4 bytes)

No ACKs or NACKs -- frames which fail CRC are simply discarded.


Collision handling

The method used is called Binary Exponential Backoff

The time after completion of one frame is divided into contention slots of length 2t, where t is the maximum propagation delay on the line.

After a collision, contenders wait 0 or 1 (chosen randomly) slot times before retransmitting.

After a second collision, contenders wait 0, 1, 2, or 3 slot times.

and so on.  After each collision, the range of wait times doubles, up to 210 = 1024 slot times.

After 16 collisions, the data link layer simply reports an error to the upper layer.

This method is a tradeoff between minimizing wait delays and minimizing collisions.

Token Ring LANs

An alternative technology for LANs is the Token Ring developed by IBM.

How it works:  A special bit pattern three bytes long called the "token" is passed around the ring.

When a station wants to transmit, it must
  • Wait for arrival of the token.  Seize it and remove it from the ring.
  • Transmit frame (first three bytes are the same as the token with one bit altered)
  • When the last bit of the transmitted frame is received, regenerate the token and switch to listen mode
Since there is only one token, only one station can send a message at any given time.  This solves the channel access problem.

In listen mode, an interface simply copies its input to its output.  The token or a message circulates around the ring.  The destination interface saves it to pass it to the destination station.

Basic characteristics:
  • Digital encoding
  • Shielded twisted pair
  • 1, 4, or 16 Mbps.

Problem:  What if a cable breaks and the ring is broken?
Solution:  Use "wire center", making the ring a "star-shaped" ring.  The wire center detects a missing station and removes it from the ring.

Problem:  What if token is lost?
Solution:  Need a "monitor" station.  If it times out, without getting the token, it drains the ring and starts a new token.

Problem:  What if the monitor goes down?
Solution:  Need a protocol to choose a new one.

Performance

  • At low load, a sender needs to wait for token.
  • At high load, becomes round robin, so no starvation is possible  (Ethernet is random)

Frame structure

SD - Starting delimiter - uses otherwise illegal pattern (Manchester encoding)
AC - Access control
FC - Frame control
Destination address (2 or 6 bytes)
Source address (2 or 6 bytes)
Data (no limit)
Checksum (4 bytes)

ED  ending delimiter
FS  Frame status




 

Powered by Plone CMS, the Open Source Content Management System

This site conforms to the following standards: