Re: [Make-wifi-fast] SmallNetBuilder article: Does OFDMA Really Work?

["Toke Høiland-Jørgensen" <toke@redhat.com>]

Tim Higgins <tim@smallnetbuilder.com> writes:

> Thanks for the additional insights, Bob. How do you measure TCP connects?
>
> Does Dave or anyone else on the bufferbloat team want to comment on
> Bob's comment that latency testing under "heavy traffic" isn't ideal?

Well, it depends on what you want to measure. Loading the link with
heavy traffic is a good way to show the worst-case behaviour of the
system, as that will fill up the buffers and expose any latent
bufferbloat. Which, as Bob points out, will tend to drown out any other
source of latency, at least if all the queues are dumb FIFOs.

However, if you want to specifically study, say, the media access
latencies of the WiFi link, drowning it out with the order-of-magnitude
higher latencies of bloat in the layers above is obviously going to
obscure the signal somewhat. Which I think was basically Bob's point
about "testing under heavy traffic"?

> My impression is that the rtt_fair_var test I used in the article and
> other RRUL-related Flent tests fully load the connection under test.
> Am I incorrect?

Yeah, the RRUL test (and friends) are specifically designed to load up
the link to show the worst-case latency behaviour, including any
bufferbloat. And, as per the above, as long as the system you're testing
still has unresolved bloat issues, well, that is what you're going to be
seeing most of... :)

-Toke