[Bloat] 250ms at 100Mbit fiber/cable

[Bloat] 250ms at 100Mbit fiber/cable

[Dave Taht]

I find the data cluster over here at 250ms/100Mbit kind of worrisome:

https://www.dslreports.com/speedtest/results/bufferbloat?up=1


-- 
I tried to build a better future, a few times:
https://wayforward.archive.org/?site=https%3A%2F%2Fwww.icei.org

Dave Täht CEO, TekLibre, LLC

Re: [Bloat] 250ms at 100Mbit fiber/cable

[Matt Taggart]

On 11/17/21 9:28 AM, Dave Taht wrote:
> I find the data cluster over here at 250ms/100Mbit kind of worrisome:
> 
> https://www.dslreports.com/speedtest/results/bufferbloat?up=1

That graph is amazing! Things I noticed:
* clear division between technologies in bandwidth
* the log scales are a great way to visualize things
* you can make out the different bandwidth "tiers" that cable providers 
offer
* fiber doesn't seem to show such tiers as much
* DSL really drops off around 12mbit as expected, but there are some 
well above that. User error when running the test?

I really like best/worst lists at the bottom.

How could this graph/data be used to measure progress on the bufferbloat 
front? It would be neat to have an animation over time and hopefully 
watch the points drift to the left.

What if the graph had the ability to toggle data sets on/off:
* by technology
* by ISP
* by region

Thanks for pointing it out

-- 
Matt Taggart
matt@lackof.org

Re: [Bloat] 250ms at 100Mbit fiber/cable

[Sebastian Moeller]

Hi Matt,


> On Nov 17, 2021, at 19:04, Matt Taggart <matt@lackof.org> wrote:
> 
> On 11/17/21 9:28 AM, Dave Taht wrote:
>> I find the data cluster over here at 250ms/100Mbit kind of worrisome:
>> https://www.dslreports.com/speedtest/results/bufferbloat?up=1
> 
> That graph is amazing! Things I noticed:
> * clear division between technologies in bandwidth

	Sampling issue, over here in Germany docsis/cable upload tops out at 50 Mbps, while VDSL2 at 40Mbps, at least on the uplink plot...


> * the log scales are a great way to visualize things

	Yes and no; what they offer on overview, they lack in detail... the difference from 250ms delay to 1 second seems smaler than from 30ms to 250ms, yet I am sure internt with 250ms latency (while painful) is leaps and bounds more usable than at 1sec....

> * you can make out the different bandwidth "tiers" that cable providers offer

	Less clear in the download scatter plot at https://www.dslreports.com/speedtest/results/bufferbloat, no? Only 100Mbps stands out.

> * fiber doesn't seem to show such tiers as much

	Funny the biggest fiber ISP in Germany sells fiber with similar rate plans as DSL (50/10, 100/40, 250/40, 500/100, 1000/200) so tiering should show up there as well, and if you clock on the cable legend text (to blend cabe samples out) you see the 250ms/100Mbps blob that Dave mentioned in the fiber results as well.

As far as I can tell the classification really comes from the button users pressed when running a test, so these categories will not be 100% pure. Especially ince the different test profiles use different numbers of flows and for most VDSL2 links the cable or fiber preset gives better throughput results (as the 4? flows in the DSL preset often ca not reliably saturate links >> 12 Mbps).


> * DSL really drops off around 12mbit as expected, but there are some well above that. User error when running the test?

	See above, in some regions/markets DSL variants offer speeds well above 12 Mbos, e.g. VDSL2 reaching 250/40, of G.fast 500/500+, the dslreports data set is not country specific.


> 
> I really like best/worst lists at the bottom.
> 
> How could this graph/data be used to measure progress on the bufferbloat front? It would be neat to have an animation over time and hopefully watch the points drift to the left.

	+1 that would be a great animation to have/show!

> 
> What if the graph had the ability to toggle data sets on/off:
> * by technology

	That you can already do, just click on the respective text in the legend and the symbol will grey out and the respective samples disappear. 

> * by ISP
> * by region
> 

	https://www.dslreports.com/speedtest/results offers by country and by ISP, but then does not show the nice log/log throughput over inferred buffer size plot for these sub groupings.

Best Regards
	Sebastian


> Thanks for pointing it out
> 
> -- 
> Matt Taggart
> matt@lackof.org
> _______________________________________________
> Bloat mailing list
> Bloat@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/bloat