- Recall the
Ethernet CSMA/CD protocol:
Observe that:
- The backoff protocol is executed when a node have (already) detected a collision !!!!
Therefore:
- Multiple nodes are
ready to
transmit !!!
(One transmission can never result in a collission.....)
- Goal
of the Backoff protocol:
- Re-schedule
the transmissions of the
collided nodes
so that:
- the likelihood (chance) of subsequent collisions is minimized.
- Re-schedule
the transmissions of the
collided nodes
so that:
- A simple-minded backoff protocol:
- Pick a
random number r
between
[0..(N−1)]
(N is
fixed)
- Wait r × τ time
Flow chart:
- Pick a
random number r
between
[0..(N−1)]
(N is
fixed)
- Weakness of the
simple-minded backoff algorithm:
- If N
is small,
and the system
is very busy
(= many nodes have collided),
then:
- It will be very likely that
more than one node
will select the
same random number
- Result:
- frequent collisions
- It will be very likely that
more than one node
will select the
same random number
- On the other hand,
if N
is large, but
the system is
lightly loaded,
then:
- Nodes may select
an
large
random value
Result:
- The node will back off for a long time (for nothing)....
- Result:
- Wasting of bandwidth.....
- Nodes may select
an
large
random value
- If N
is small,
and the system
is very busy
- Conclusion:
- We need a dynamic backoff algorithm that can adjust to different network operating conditions !!!!
- Adaptive back off:
- Allow collided nodes to
pick small back off periods
in earlier
re-transmission attempts
- And pick larger back off periods when there are more re-transmission attempts
- Allow collided nodes to
pick small back off periods
in earlier
re-transmission attempts
- The exponential backoff algorithm
flow chart:
- Explanation:
- When a frame is transmitted
for the first time:
Reset k = 0
- If there was a collision,
the
first
retransmission attempt will use:
k = k + 1 = 0 + 1 = 1 N = 2k - 1 = 21 - 1 = 2 - 1 = 1
For the first re-transmission attempt, the node will picks a random number r from the range:
- r ∈ [0, 1]
This range is suitable when there are very few (about 2) nodes involved in the collision
- It the node
is involved in
another collision
while trying to
transmit the
same frame,
then:
k = k + 1 = 1 + 1 = 2 N = 2k - 1 = 22 - 1 = 4 - 1 = 3
So the second retransmission attempt for the same frame will use a random number r from the range:
- r ∈ [0..3]
The node is adjusting to accomodate for a larger number of nodes (i.e., a heavier loaded situation)
- If there is another collision,
the
third
retransmission attempt
of the same frame will use:
- r ∈ [0..7]
- And so on...
- When a frame is transmitted
for the first time:
- Fact:
- One Ethernet network can support at most 500 nodes
Therefore:
-
The maximum range from which the
random numbers are selected is:
- [0..210-1] = [0..1023]
- The final version of
the Ethernet Medium Access Control (MAC)
algorithm:
