[Bloat] [Make-wifi-fast] The most wonderful video ever about bufferbloat

Sebastian Moeller moeller0 at gmx.de
Sat Oct 22 15:47:48 EDT 2022


Hi Matt,


> On Oct 22, 2022, at 20:37, Matt Taggart via Bloat <bloat at lists.bufferbloat.net> wrote:
> 
> On 10/17/22 19:58, David Lang via Bloat wrote:
> 
>> actually, fair queueing is more like the '15 items or less' lanes to speed through the people doing simple things rather than having them wait behind the mother of 7 doing their monthly shopping.
> 
> Sort of, but CoS/QoS could be described that way too? But at least it's based on the quantity of items, CoS/QoS would be more like "moms with kids go in this line, business execs go in this line, everybody else in this other line" and "we have 4 checkers available for the business exec line, 2 for the mom&kid line, and the remaining for everyone else".
> 
> Maybe the best is: I'm standing in long line with a huge cart of groceries and there is a person behind me with only 3 things, so I tell them to go ahead of me so their wait time isn't even worse. But usually this happens because the people in line make it happen, the checkers default to FIFO out of "fairness" (for some value of fair).
> 
> Maybe car-pool lanes are a good analogy, but some of them are HOV and some of them allow you to pay to use...

[SM] None of the real life models are all that well fitting for the problem, information simply is different from tangible objects in that dropping and resending on failure are less problematic. But that means that all "intuitive" analogies are out and trying to explain internet congestion and remedies with telegrams or even letters does not really help all that much (and in both cases dropping was rare and considered a failure not part of normal operations as far as I can see).
Different queueing strategies have analogies (think a road/bridge pay terminal which might be single queue fifo or might offer multiple parallel booths) but AQM really is hard to get an analogy for... probably because most analogies put us into the mindset of an individual packet, while AQMs operate on the assumption of packet trains that will respond to signaling. To conjure something up this is a bit like going to the sports-ball game every Sunday with a special ticket that allows unlimited +1s if the stadion is not yet capacity and them telling you the time you (this is ECN) come with 10 friends to try with fewer friends next time... or simply only letting say 5 in (dropping the rest) so next time you try only with 5... that is all but intuitive.


Regards
	Sebastian



> 
> Also couldn't resist :)
> 
> https://www.youtube.com/watch?v=RGWiTvYZR_w
> 
> -- 
> Matt Taggart
> matt at lackof.org
> _______________________________________________
> Bloat mailing list
> Bloat at lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/bloat



More information about the Bloat mailing list