Hacker Newsnew | past | comments | ask | show | jobs | submitlogin
Flent: The FLExible Network Tester (flent.org)
81 points by wallflower on Jan 28, 2024 | hide | past | favorite | 10 comments


It is not in the same league as Flent, but I really like https://speed.cloudflare.com as an in-browser internet speed test. It's backed by a very densely deployed CDN, it tests both loaded and unloaded latency, and it reports jitter and draws candlestick plots.


> https://speed.cloudflare.com

Love this for super quick no brainer data. I used to use speedtest.net, by Ookla, back in the day, but this is so much better in so many ways and provides so much more data.


What it has that virtually every speedtest lacks is a packet loss test that floods 1000 packets and checks if it receives a reply. The bad news is sometimes it doesn't work, and sometimes the results are incorrect (indicating packet loss when generally there is no loss which might be more of an indicator that something is wrong with CF).


I keep getting surprised by how good it is. It's a very exhaustive test, everything has an explanation, and it just... works.

Just about the only thing it is missing is IPv4/IPv6 capability (Can I connect to 4-only? 6-only? Dualstack?), and perhaps some kind of traceroute.


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


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-testin...


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/


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?


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...


Dslreport’s speed test also does bufferbloat testing.




Consider applying for YC's Winter 2026 batch! Applications are open till Nov 10

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: