Bufferbloat?

What's this new speed test actually telling me about my Vivid 200 connection?

213 Mbps down, 12.5 Mbps up - this looks pretty good.

Your latency: 18 ms - seems fine for my location, could be better but is good enough for LoL

Bufferbloat v D ^ F - What ???

The explanation seems rather biased against certain network technologies:

"Bufferbloat is a measure of how delayed packets are due to excessive buffering in broadband routers and other locations on the Internet.

A is good, F is bad and is only something to be worried about if you upload a lot while also downloading or vice versa."

I'm reliably informed by my son that our connection is actually very good in terms of latency/lag when playing LoL regardless of what others are doing in the household.

Question...does the son game while downloading a 50GB file that is arriving at the full 213 Mbps? Or do they time these to be at other times...my bet is the latter and this is because of buffer bloat

General web browsing and video streaming, even upload won't cause an issue, but buffer bloat is all about the delays introduced when you have saturated either the upload or download sides.

As for being 'biased against certain network technologies' I'd like to hear why you believe that to be the case, since there is no bias in the code or theory.

The buffer bloat and quality metric which is not reported yet to individual speed test users (but used in monthly round-up) are mechanisms to help expand how broadband is measured and extending beyond the game for just ever higher numbers.

Knowing who to show extra metrics to and who not to is impossible to predict, am planning to add an 'advanced' tab giving more detail to those who want it.

In reply to a post by MrSaffron:

Question...does the son game while downloading a 50GB file that is arriving at the full 213 Mbps? Or do they time these to be at other times...my bet is the latter and this is because of buffer bloat

He doesn't but if I happen to kick off a steam download while he is playing he doesn't notice, and LoL is very sensitive. I also have two daughters that are almost permanently attached to Netflix/YouTube. Even with all the bits going on with Nest, other cloud and RDP connections I don't often saturate the upstream but I've never had him complain when I have done the odd 4K upload. Hence my query as to what it actually means when it says my connection is bad?

In reply to a post by MrSaffron:

As for being 'biased against certain network technologies' I'd like to hear why you believe that to be the case, since there is no bias in the code or theory.

We all know that, due to the way it works, cable needs more buffering which is why it suffers from higher jitter. My guess is that this is also feeding into the 'Bufferbloat' assessment and leads to showing a non-DSL connection that appears to work very well in all real world scenarios as a 'bad' connection (somewhat akin to the focus on single-threaded speed which is also affected by cable's required buffering).

Buffering is nothing to do with single threaded speed issues, RWIN may mean it takes longer to reach full speed with a single thread, but given wider feedback from those in areas where they have issues I see no reason to drop single threaded testing.

Good to hear that saturating your downstream with a steam download (I presume it runs at connection speed) is not affecting game play. Maybe there is some UDP prioritisation at play somewhere.

It's good to hear that your latency/lag doesn't get bad during downloads. But the DSLReports.com/speedtest measures latency during transfers of both directions. I suspect the D-F grade comes from uploads.

Another way to test is to start a ping test to google.com and then run any speed test. Look to see if the latency increases during either phase of the test.

See "What can I do about Bufferbloat?" at https://www.bufferbloat.net/projects/bloat/wiki/What... for more info

No, the D is the grade for download, the F is for upload. The new speedtest shows these with little graphic arrows which I substituted to v and ^.

Both imply I have a 'bad' connection though I haven't seen any practical evidence. It won't be long before someone is bashing VM about this just like they do on the single-thread issue.

your reasoning is a bit like saying speedtests are unfair on dialup tech
bufferbloat only matters if you care about usability of connection during periods when the connection is saturated, if that is of no concern then dont stress over it.
But be aware tho in case of VM bufferbloat can also occur with just the headend saturated not your own connection.

Edited by Chrysalis (Sat 13-May-17 07:17:28)

I am absolutely certain that providers would love me more if I dropped all diagnostics and just showed a 75th percentile result for download and make the upload testing an advanced option.

In reply to a post by MrSaffron:

I am absolutely certain that providers would love me more if I dropped all diagnostics and just showed a 75th percentile result for download and make the upload testing an advanced option.

They might we wouldn't. We rely on your scientific approach.

Isn't that just what fast.com does?!

In reply to a post by Daemon66:

Both imply I have a 'bad' connection though I haven't seen any practical evidence. It won't be long before someone is bashing VM about this just like they do on the single-thread issue.

One reason NOT to read too much into test results. If it's running OK, then what is the point of testing connection?
Yes, if you are getting problems. Then testing can help prove a point. But if nothing is wrong in "Real life usage". Then often such as this tread, cause nothing other than confusion.