Whenever an end station MAC receives a transmit-frame
request with the accompanying address and data information from the LLC sublayer,
the MAC begins the transmission sequence by transferring the LLC information
into the MAC frame buffer.
� The
preamble and start-of-frame delimiter are inserted in the PRE and SOF fields.
�The
destination and source addresses are inserted into the address fields.
�The
LLC data bytes are counted, and the number of bytes is inserted into the
Length/Type field.
�The
LLC data bytes are inserted into the Data field. If the number of LLC data bytes
is less than 46, a pad is added to bring the Data field length up to 46.
�An
FCS value is generated over the DA, SA, Length/Type, and Data fields and is
appended to the end of the Data field.
After the frame is assembled, actual frame transmission
will depend on whether the MAC is operating in half-duplex or full-duplex mode.
The IEEE 802.3 standard currently requires that all
Ethernet MACs support half-duplex operation, in which the MAC can be either
transmitting or receiving a frame, but it cannot be doing both simultaneously.
Full-duplex operation is an optional MAC capability that allows the MAC to
transmit and receive frames simultaneously.
Half-Duplex Transmission�The CSMA/CD Access Method
The CSMA/CD protocol was originally developed as a means by
which two or more stations could share a common media in a switch-less
environment when the protocol does not require central arbitration, access
tokens, or assigned time slots to indicate when a station will be allowed to
transmit. Each Ethernet MAC determines for itself when it will be allowed to
send a frame.
The CSMA/CD access rules are summarized by the protocol's
acronym:
�Carrier
sense�Each station continuously listens for traffic on the medium to
determine when gaps between frame transmissions occur.
�Multiple
access�Stations may begin transmitting any time they detect that the
network is quiet (there is no traffic).
�Collision
detect�If two or more stations in the same CSMA/CD network (collision
domain) begin transmitting at approximately the same time, the bit streams from
the transmitting stations will interfere (collide) with each other, and both
transmissions will be unreadable. If that happens, each transmitting station
must be capable of detecting that a collision has occurred before it has
finished sending its frame.
Each must stop transmitting as soon as it has detected the collision and then
must wait a quasirandom length of time (determined by a back-off algorithm)
before attempting to retransmit the frame.
The worst-case situation occurs when the two most-distant
stations on the network both need to send a frame and when the second station
does not begin transmitting until just before the frame from the first station
arrives. The collision will be detected almost immediately by the second
station, but it will not be detected by the first station until the corrupted
signal has propagated all the way back to that station. The maximum time that is
required to detect a collision (the collision window, or "slot time") is
approximately equal to twice the signal propagation time between the two
most-distant stations on the network.
This means that both the minimum frame length and the
maximum collision diameter are directly related to the slot time. Longer minimum
frame lengths translate to longer slot times and larger collision diameters;
shorter minimum frame lengths correspond to shorter slot times and smaller
collision diameters.
The trade-off was between the need to reduce the impact of
collision recovery and the need for network diameters to be large enough to
accommodate reasonable network sizes. The compromise was to choose a maximum
network diameter (about 2500 meters) and then to set the minimum frame length
long enough to ensure detection of all worst-case collisions.
| 
