- Reminder:
- The Distance Vector Algorithm is
a gossip-based algorithm
- Nodes exchange distance vector messages with its immediate neighbors to perform the computation
- The Distance Vector Algorithm is
a gossip-based algorithm
- Vector:
- Vector = a
row or
column of
numbers
Example:
1 5 2
- Distance vector = a vector of distance values
- Vector = a
row or
column of
numbers
- Structure of
a distance vector message:
Distance vector message sent by some node N: To Distance ----+---------- A 4 (meaning: distance(N, A) = 4 B * (meaning: distance(N, B) = infinite) .. .. .. ..Example:
- Node A is
aware
of its distance to
B and D:
- Node A will
send the following
distance vector message
to its
immediate neighbors:
- Node A is
aware
of its distance to
B and D:
- Trigger:
- Whenever the
distance
to some destination
changes:
- The node must send a distance vector message (to all its immediate neighbors)
Comment:
- A node does
not need to
send the distance vector
update message
immediately:
- A node can collect the updates for some period of time and
- Then send all the updates in one distance vector message
- Whenever the
distance
to some destination
changes:
- Dictatical note:
- The distance vector message
only need to contain the
changed distance values
But:
- In my examples,
I list the distances to
all nodes
(That's because its easier to cut-and-paste and change some values than write out a new vector each time !!!)
- In my examples,
I list the distances to
all nodes
- The distance vector message
only need to contain the
changed distance values
- Suppose a node receives
a distance vector message:
From: NNode Distance vector message: Dest | Distance --------+------------- A | dA B | dB ... | ... - Update procedure:
Visual help: distance vector message
/* ---------------------------------------------
Update distances to "Dest" going "Via NNode"
--------------------------------------------- */
for ( each entry (Dest, Dist) in the distance vector message ) do
{
Distance table:
Dest | Via X ... Via NNode ... Via Y
-------+------------------------------------
A | ... ... d(Me,NNode) + dA ... ...
B | ... ... d(Me,NNode) + dB ... ...
.......
Visual help:
}
/* -----------------------------------------------------
Find updated min. distance to "Dest" after an update
----------------------------------------------------- */
for ( each entry (Dest, Dist) in the distance vector message ) do
{
d(Me, Dest) = minall NNode x ( Via x | d(myself, Dest) );
Visual help:
/* -----------------------------------------------------
Propagate a change to your neighbors
---------------------------------------------------- */
if ( d(Me, Dest) has changed )
{
Schedule update message "Me, ToNode, d(myself, Dest)"
Visual help:
}
}
- Recall:
- To reduce the
number of distance vector messages,
the updates
can be collected for
a period of time
- At the end of the collection period, all distance updates are sent in one message
- To reduce the
number of distance vector messages,
the updates
can be collected for
a period of time