(rev. 01/20/2010)
Notes On Chapter Fourteen
-- The IEEE MAC Sub-Layer
- 14.1 Introduction
- Multi-access Protocols
- Static and Dynamic Channel Allocation
- 14.2 A Taxonomy of Mechanisms for Multi-Access
- Three approaches by which computers can coordinate access to a shared
medium are:
- Controlled Access Protocols - for example:
- Reservation
- Polling
- Token Passing
- Random Access Protocols - for example:
- Channelization Protocols - for example:
- 14.3 Static and Dynamic Channel Allocation
- An FDM system can assign a channel for the exclusive use of a
communicating pair of computers.
- In some cases a static assignment will suffice.
- In many cases the pairs communicating will vary over time, so
channels need to be allocated and deallocated dynamically.
- 14.4 Channelization Protocols
- Examples:
- FDMA - Frequency Division Multi-Access - for example:
- 'New station' uses reserved control channel to make a
request to a 'central controller'
- Central controller chooses an unused channel and
assigns it to the station
- TDMA - Time Division Multi-Access - for example:
- A station is assigned a time slot during which it is
allowed to send
- This could be a round-robin scheme -- so, for example,
the station might be allowed to send every 3rd packet among
a set of 25 stations.
- CDMA - Code Division Multi-Access
- All senders may send at the same time - they
send "mutually orthogonal" data which is multiplexed by
summing. This is comparable to interleaving at the bit
level.
- 14.5 Controlled Access Protocols
- 14.5.1 Polling
- Centralized controller polls 'stations',giving them an
opportunity to send one packet
- selection may be simple round robin or some priority scheme.
- the controller sends a polling message to a station and the
station responds, either by sending a packet, or 'passing'.
- 14.5.2 Reservation
- Stations send messages in rounds.
- Typically for a period of time, stations send
messages to a central controller saying whether they will
have a need to send a message in the next round.
- Then the controller transmits a list of stations that will be
transmitting in the next round.
- Stations then use the list to determine when to send during the
next round.
- Variation: there is a separate channel for controller to use to
gather reservations.
- This technique is often used with satellite transmission.
- 14.5.3 Token Passing
- There is a special control message called the token.
- The token is passed from station to station according to some
preassigned 'circular order'.
- When a station has the token, it is allowed to transmit a
message.
- The station passes the token immediately, or right after sending
its message.
- 14.6 Random Access Protocols
- 14.6.1 ALOHA
- A protocol used in an early data network in Hawaii.
- The medium was radio frequency
- ALOHA pioneered 'random access techniques'
- A master central station with a powerful transmitter is
surrounded by 'slave stations' with weaker transmitters.
- Transmissions of the master can reach all slaves, and
transmissions of all slaves can reach the master.
- Transmissions of most slaves can't reach all the other slaves.
- The network employs two different frequencies - an
outbound frequency for the master to send messages to
slaves, and an inbound frequency for slaves to send
messages to the master.
- A slave that wants to send a packet to another slave, transmits
it to the master on the inbound frequency. The master then
resends the packet on the outbound frequency. Both the intended
receiver and the sender get a copy of the packet sent by the
master. The copy that is received by the original sender serves
as an acknowledgment that the packet arrived safely.
- Two slaves might send a packet at the same time. If so they
interfere with each other (collide), are unintelligible
to the master, and are NOT resent by the master on the outbound
frequency.
- If a slave sends a packet, and does not soon receive a copy of
the packet back on the outbound frequency, then the slave will
retransmit the packet, after a random delay.
- Because retransmission delays are randomly chosen, there is a
low probability that the two senders that collided before will
collide again.
- If the network is heavily utilized collisions can happen
frequently despite these measures.
- 14.6.2 CSMA/CD
- The original Ethernet employs a single long cable to which all
computers connect and transmit.
- Ethernet Carrier Sense Multi-Access with Collision Detection
(CSMA/CD) is similar to ALOHA but utilizes several innovations:
- The "two-frequency idea" was dropped.
- An NIC "carrier senses" and refrain from attempting to send
while a transmission is in progress (this improves
utilization).
- Stations monitor the ether while sending, and abort
transmission if a collision is detected.
- After detecting a collision, an NIC utilizes a binary
exponential backoff strategy:
- wait a random time (bounded by some constant d) and
attempts to retransmit.
- If a second collision occurs, abort again and picks
another random delay, this time bounded by 2d.
- If collisions keep recurring while trying to send the
same packet, keep doubling the maximum delay until
success (or after trying some large number of times,
give up.)
- 14.6.3 CSMA/CA
- One of the things that allows CSMA/CD to work on an Ethernet
is the fact that all NIC's are able to receive each other's
transmissions.
- This is not true in the typical wireless LAN.
- C1 --------- C2 ------------ C3
Suppose (C1,C2) and (C2,C3) are in range, but (C1,C3) are not.
Suppose further that C1 and C3 send at the same time, intending
that C2 be the receiver. The packets will collide at C2, but
the collision will not be detected by C1 or C3.
- There is an alternative called Carrier Sense Multiple Access
with Collision Avoidance.
- When C3 wants to send to C2 it sends a very brief message
announcing that fact.
- Upon receiving that message C2 sends out a brief message saying
it is about to receive a packet from C3.
- Every computer within range of either C3 or C2 receives one or
both of these short messages, and refrains from sending for a
long enough time so that C3 can send its packet to C2 without
interference.
- If a control message from C3 collides at C2 with a control
message from C1, then much like with ALOHA, C2 will NOT reply
and C1 and C3 will apply a random backoff scheme until one of
them succeeds in getting a control message through to C2.