Latency grade support

[mpenning]

Would it make sense for flent to provide a latency grade?

I personally would love to have this instead of having to go off-net to get a latency grade.

If flent already does this, then please clue me in for which test to run.

[tohojo]

Flent does not suppose any particular scenario or test target (you can use it to test your LAN or your satellite link just fine, that all depends on what you point the test at), so having a "grade" does not really make sense, I think. You can look at then summary numbers (that Flent outputs on the command line by default) and draw your own conclusion as to what constitutes "good" latency in the scenario you are testing...

[mpenning]

Flent does not suppose any particular scenario or test target (you can use it to test your LAN or your satellite link just fine, that all depends on what you point the test at), so having a "grade" does not really make sense, I think. You can look at then summary numbers (that Flent outputs on the command line by default) and draw your own conclusion as to what constitutes "good" latency in the scenario you are testing...

My premise for proposing a bufferbloat grade:

• Providers somehow support a bufferbloat grade without knowing what you are testing on (could be Satellite or terrestrial)
• You can input any numbers you want (such as "expected average latency")
• Flent has built a massively-useful framework around testing latency, but the options are so wide it's helpful just to get a rough "thumbs-up" or "thumbs-down" kind of result; this is why I ask about a bufferbloat grade

Flent obviously does not have to support this, but it would be useful to have an open-source implementation of a latency grade algorithm (written by experts).

[tohojo]

Mike Pennington ***@***.***> writes:

> Flent does not suppose any particular scenario or test target (you can use it to test your LAN or your satellite link just fine, that all depends on what you point the test at), so having a "grade" does not really make sense, I think. You can look at then summary numbers (that Flent outputs on the command line by default) and draw your own conclusion as to what constitutes "good" latency in the scenario you are testing... My premise for proposing a bufferbloat grade: - Providers somehow support a bufferbloat grade without knowing what you are testing on (could be Satellite or terrestrial) - You can input any numbers you want (such as "expected average latency") - Flent has built a massively-useful framework around testing latency, but the options are so wide it's helpful just to get a rough "thumbs-up" or "thumbs-down" kind of result; this is why I ask about a bufferbloat grade Flent obviously does not have to support this, but it would be useful to have an open-source implementation of a latency grade algorithm (written by experts).

Right, I understand your reasoning, and it would not be totally unreasonable to have an output mode that implements this (a Formatter class would probably be the easiest way to do this). I do not currently have any spare cycles to work on this, but if someone contributes an implementation I am happy to look at it :)

[dtaht]

Getting to where we could postprocess a grade that used waveformś' algorithm or others would be good. One day.