views:

48

answers:

4

considering that metric is delay in distance vector routing algorithm, is it possible that delay from router A to B is different from router B to A. if yes, under which conditions??

thanks.

A: 

Are you assuming both hypothetical circumstances run at the exact same time? If not I suppose there could be a spike on the traffic for one of the routers at any given time that bogs down your 'wanted' traffic.

Ian Jacobs
+1  A: 

The algorithm assumes the graph is bidirectional. Of course, it's possible for the delays to be different in each direction in practice: for example, if B is transmitting heavily to A, then traffic from A to B is likely to be faster than from B to A, since traffic from B will have to get in line at the end of a queue.

WhirlWind
+1  A: 

Delay and metric are two different things.

  • Delay is the time it takes for a packet to traverse the network. If a link is heavily utilized in one direction and there is some kind of buffering device (such as a switch) on the link you might have different delays in the network traffic depending on direction.
  • Metrics are values associated with entries in a routing table that indicates "costs" of different routes. If A and B have static routing entries they can definitely be configured with different metrics for each direction of the same link.
Anders Abel
A: 

Certainly this is possible, but to give you more details you probably need to be more specific with the question.

With regards to your specific question about Metrics and Distance Vector routing algorithms, yes, A can be configured to think that B is further away than B thinks A is, although as mentioned by one of the other answers, that doesn't necessarily mean the delay is different although it may in fact be.

In practice though, there are lots of questions to consider:

Is router A adjacent to router B? If not, then you certainly could have different delays because inbound packets may take a different path than outbound packets.

If they are adjacent, what kind of connectivity do they have? Are they the same kind of router? Imagine a router at the end of an aysymmetric DSL line. Of course the propagation delay wouldn't be aysymmetric, but delay could be higher in one direction as a result of traffic congestion. (This scenario also gives a concrete example of why you might want A to think the link to B has a higher cost than B thinks the link to A has.)

In practice, the definition of delay makes a big difference too. Are you thinking of just the computed cost? Or just propagation delay? Or just the link cost? If router B is sending more traffic than router A, it may take longer for responding packets from B to A to be processed by B than A takes when sending the packets (the same may apply for intermediary switches, especially in the case of things like multicast packets--some routers and/or switches take longer to process multicast and other "special" packets). So in this scenario the actual delay may be different, but the cost the DVP is using thinks it is the same.

Hope this answer helps. Good luck,

--jed

Jed Daniels
This question pertains specifically to an algorithm.
WhirlWind
Thanks @WhirlWind, I've corrected the answer to more accurately reflect what was being asked. I read the question too quickly the first time.
Jed Daniels