3 ms·

Dave Taht's bufferbloat project could have gotten way more actual non-technical user data had Flent supported Windows. At least fast.com and Waveform's bufferbl

Dave Taht's bufferbloat project could have gotten way more actual non-technical user data had Flent supported Windows. At least fast.com and Waveform's bufferbloat test [1] exist now. [1]: https://www.waveform.com/tools/bufferbloat https://www.waveform.com/tools/bufferbloat

dtaht 3y ago
yes, had we found a way to support windows, we could have (and could still do) better. Regrettably we never did find a way - or funding - to drive tests through windows (wsl works). I still (as do many other folk in the Bufferbloat world) rely on flent to drive all kinds of tests. nothing compares to flent's ability to compare multiple network tests in so many ways. https://blog.cerowrt.org/post/flaws_in_flent/ https://blog.cerowrt.org/post/flaws_in_flent/
dtaht 3y ago
I have been steering folk at this new test - called crusader - as it captures the essence of the baseline rrul flent test testing both up + down and latency at the same time - and as it is written in rust, runs on everything. https://github.com/Zoxc/crusader/releases/tag/v0.0.10-testing https://github.com/Zoxc/crusader/releases/tag/v0.0.10-testin...
evulhotdog 3y ago
Dslreport’s speed test also does bufferbloat testing.
BobbyTables2 3y ago
I feel like waveform should be reported to the FTC. They basically state that the solution to their test failing is a new WiFi router. That is not going to do much when the test is run from a wired connection! I can change my “grade” from F to A simply by switching to the Vegas TCP congestion algorithm. Maybe they were formerly car salespeople?
- dtaht 3y ago
  Switching to a router that supports SQM (ideally, CAKE) on the wan interface is the suggested solution. fq_codel on the wifi, helps a lot too, but first up is fixing the ISP connection. Or an ISP can install libreqos, or preseem, or paraqum... Ironically I ran tcp vegas and SFQ, by default, since roughly 2002, since I cared about voip quality in particular. The rest of the world, didn't. Thus fq_codel was born...

dtaht 3y ago

yes, had we found a way to support windows, we could have (and could still do) better. Regrettably we never did find a way - or funding - to drive tests through windows (wsl works). I still (as do many other folk in the Bufferbloat world) rely on flent to drive all kinds of tests. nothing compares to flent's ability to compare multiple network tests in so many ways. https://blog.cerowrt.org/post/flaws_in_flent/ https://blog.cerowrt.org/post/flaws_in_flent/

dtaht 3y ago

I have been steering folk at this new test - called crusader - as it captures the essence of the baseline rrul flent test testing both up + down and latency at the same time - and as it is written in rust, runs on everything. https://github.com/Zoxc/crusader/releases/tag/v0.0.10-testing https://github.com/Zoxc/crusader/releases/tag/v0.0.10-testin...

evulhotdog 3y ago

Dslreport’s speed test also does bufferbloat testing.

BobbyTables2 3y ago

I feel like waveform should be reported to the FTC. They basically state that the solution to their test failing is a new WiFi router. That is not going to do much when the test is run from a wired connection! I can change my “grade” from F to A simply by switching to the Vegas TCP congestion algorithm. Maybe they were formerly car salespeople?

dtaht 3y ago
Switching to a router that supports SQM (ideally, CAKE) on the wan interface is the suggested solution. fq_codel on the wifi, helps a lot too, but first up is fixing the ISP connection. Or an ISP can install libreqos, or preseem, or paraqum... Ironically I ran tcp vegas and SFQ, by default, since roughly 2002, since I cared about voip quality in particular. The rest of the world, didn't. Thus fq_codel was born...