Is it possible to use ATM on fiber links?

If so, what would be the overhead and the correct mpu in SQM configurations?

Yes, you can use ATM over fiber links, and in the past that has been deployed, see e.g. here:
https://en.m.wikipedia.org/wiki/Synchronous_optical_networking

Overhead and mpu then depend on the encapsulation actually in use...
If ATM/AAL5 is used you can use the following to empirically deduce the overhead:
https://github.com/moeller0/ATM_overhead_detector

Question, why would you want to do that?

The good thing is that I think my provider does this. Recently I'm using SQM for games and I don't see any difference with it disabled or enabled. Then I remembered that when the technician came to install the fiber, we asked him to cut the fiber from the old operator, however he said that the fiber was not implemented in the standard way, it was in a low cost way and there was no way to remove the cable from the old one. operator, only the operator itself could do this. After remembering this, I changed the link layer adaptation to ATM and the gaming experience was much better. The question is does what I say make any sense or is it just coincidence?

Well, if you use the github project above you should be able to detect whether your link uses ATM/AAL5. I would think this is rather unlikely, but certainly not impossible.... So, please post the results of the ATM_overhead_detector to this thread and we can ponder whether your ISP configured such a unicorn link or not :wink:

Hi, I tried to run your script but Windows says that the hrping command is not recognized. hrping only works if I run the command in its directory. Would you know why this is?

I guess you can either copy the ping_collector.bat script to the directory where the hping binary lives or add that directory to your windows path variable.
HOWEVER, please re-download ping_collector.bat from the git repository, as the old version tries to use google's 8.8.8.8 (which only honors the size request up to 68 bytes, which for ATM detection is not enough).

How do I use ATM_overhead_detector? I'm at step 4.

[main 2024-03-21T22:21:08.950Z] update#setState disabled
[main 2024-03-21T22:21:08.951Z] update#ctor - updates are disabled due to running as Admin in user setup
[main 2024-03-21T22:21:13.926Z] Extension host with pid 8340 exited with code: 0, signal: unknown

Mmh, not sure what is up, but if you share the output of the bat file, I could run the analysis...

I sent it in private chat.


IMHO this does not look at all like ATM/AAL5 is in use... (ignore the 'Quantized ATM carrier LIKELY' text, that is triggered simply by the stair fit residual being smaller than the linear fit residual, but for good measurements on real ATM/AAL5 links the ratio linear/stair is much larger that the 1:1.03 seen in your case, for a clear ATM/AAL5 example see here:

So the options are:
a) the measurement was distorted by lots of additional traffic on your link, but the link uses ATM/AAL5
b) the link does not use ATM/AAL5

I would tend to assume b), because ATM/AAL5 really is not an attractive option for an ISP in the 2020s...

My best guess is, you do not have multiple point to point fibers, but a single shared PON fiber and that meas´ns there is only one ISP actually operating the fiber and alternative ISPs offer their services via bitstream access, that is they pay the operating ISP for shuttling digital data between the end-user and the alternative ISP's network. But since there is only a single optical fiber you can not cut this when changing ISPs... (there is an optical unit connected to the fiber, often called an ONT or ONU, I bet if you google that devices name or model number it will reaveal some sort of PON technology, most likely GPON).

Probably coincidence... ATM overhead is pretty substantial (it adds around 100-100*48/53 = 9.4% to the difference between the shaper gross rate and the maximal possible throughput measured in applications like speed-tests (actually these are misnamed, they measure capacity, not speed)). So you might as well simply reduce the gross shaper rates by another ~10% to get the same effect...

So I understand, thanks for your help.

This topic was automatically closed 10 days after the last reply. New replies are no longer allowed.