Collisions metric in OLSR point to point

[adhysa...@gmail.com]

Hi all,

I want to compute the number of collisions in the wireless channel (at the nodes more specifically). So, I modifying the `/src/olsr/examples/simple-point-to-point-olsr.cc` to report the collisions. I looked at the `src/wifi/examples` and there are few codes that report collision, but I couldn't not figure out how to port that procedure to my code. I also looked at the several questions in this mailing list, but also couldn't find something useful to me yet.

Can you help please?

Thanks in advance.

[Konstantinos]

Hi,

The scenario you use for base, simple-point-to-point-olsr.cc, will not show any collisions as it is a wired network. P2P NetDevices in ns-3 are wired.

Have a look at the wireless examples 

Regards

K

[adhysa...@gmail.com]

My mistake. I was mislead by the fact that OLSR is mostly used in ad hoc networks.

Please find attached my sample code. It is a 3x3 grid with sender and receiver on opposite ends.

I added the PhyRxDrop Tracer, but the amount of dropped packets for node zero is 23.101, which is kinda high in my opinion.

PhyRXDrop - Time: 99.9915 has number of packets PhyRxDrop = 23099
PhyRXDrop - Time: 99.994 has number of packets PhyRxDrop = 23100
PhyRXDrop - Time: 99.9989 has number of packets PhyRxDrop = 23101

Ideally, I want to add the Collision test done here (https://www.nsnam.org/doxygen/wifi-phy-test_8cc_source.html), but it uses two nodes. How can I add a collisions metric?

Thanks

[Konstantinos]

Hi,

The PhtRxDrop does not mean that there is a collision. It could just be that the Rx signal (Ed threshold) is too small, or there is issue with the plcp preamble reception

Regards

K

[adhysa...@gmail.com]

Hi Konstantinos,

If packets that interfere in transmissions are higher than the Ed Threshold, it will raise the PhyRxDrop at EndReceive(), right? Can I infer that those packets are the ones I'm interested in?

You mentioned collision in your CLWPR paper. Did you end up measuring collisions?

Is there a more reliable way to track collisions? The threads I see in this mailing list are not always conclusive in this topic (including this that states: The collision detection is also difficult in real world because you don't get anything detectable when the signal destroys before the frame is received. It is later inferred from the lack of an ack.)

Thanks