dtaht

16 ms·

211.
by dtaht 5y ago
The "sparse" and "rounds" concepts also works really well as a per station wifi scheduler, where it is now the default for a plethora of chipsets in linux. Some plots of which I will forever be proud here: https:/&
212.
by dtaht 5y ago
Close. A major innovation in the fq-codel derived algorithms over former forms of FQ like DRR and SFQ is what we call the sparse flow optimization. Packets from flows that have an arrival rate lower than the departure rate of 1 quantums wor
213.
by dtaht 5y ago
Queue theory in general applies not just to supply chains, or bufferbloat, but explains a lot about how to do JIT and other forms of "traffic scheduling". Not enough people have read Len Kleinrock's wonderful books on this su
214.
Vint Cerf – Ask me anything on Feb 11(discuss.broadband.money)
2 points by dtaht 5y ago|0 comments
215.
by dtaht 5y ago
Head drop, so long as it preserves a "round" so a malignant sender cannot force all other traffic out of the queue, is great . This is what fq-codel, fq-pie, and cake do.
216.
by dtaht 5y ago
Fair queueing makes a huge difference for a large percentage of traffic, and generally makes an AQM work better by better muxing packets. I have a long list of things that can be done listed here: http://www.taht.net/~d&#x2F
217.
by dtaht 5y ago
it was without question, one of your more brief, poignant pieces.
218.
by dtaht 5y ago
The modern AQM's are based on "Time in Queue". Both pie and codel work brilliantly with pause frames, so long as BQL is also in the driver. fq+aqm (be it codel, pie or cobalt) works even better. See: https://datatr
219.
by dtaht 5y ago
I tried to leave people laughing, here: https://blog.apnic.net/2020/01/22/bufferbloat-may-be-solved-... And the online book, freely available and primarily on applying fq_codel to everything (and also sch_cak
220.
by dtaht 5y ago
"It is time to update our understanding of the primary factors directly affecting end-user Internet performance.What we have learned is that high throughput alone is not sufficient." - from the executive summary
221.
Working Latency, Explained – By the Broadband Internet Technical Advisory Group [pdf](bitag.org)
4 points by dtaht 5y ago|1 comments
222.
by dtaht 5y ago
I've given more than a few lectures on how packets and queues and aqm are supposed to work, with very physical demos, using "packets as people", such as this one: https://blog.apnic.net/2020/01/22&#x
223.
by dtaht 5y ago
by multiplexing better, you generally need to not differentiate between high and low priority. Being that I've been trying to convince people of this for years - all I can do is try to point at resources that demonstrate how we think a
224.
by dtaht 5y ago
Keeping the queues short with AQM makes every application share more fairly. Low rate applications like gaming and zoom benefit. QoS, as you describe, has been tried (see diffserv and intserv), and fell down because mostly everyone felt &qu
225.
by dtaht 5y ago
AQM technologies were described as necessary best current practice by the IETF back in 1992, and recommendations revised in 2013 https://datatracker.ietf.org/doc/rfc7567/ because the algorithm for it chosen in 199
226.
by dtaht 5y ago
sometimes... I hate how popular avery's posts are. (he's my former boss). I simply can't write like that. Or think like that. good show, ex-boss!
227.
Comcast reduced “working latency” 90% by deploying AQM during Covid(blog.apnic.net)
20 points by dtaht 5y ago|9 comments
228.
by dtaht 5y ago
I am a bit sad that what I felt was the proposal of most immediate benefit to the internet was not the one that got all the attention. That one is here: https://datatracker.ietf.org/doc/draft-schoen-intarea-unicas...
229.
by dtaht 5y ago
A very good overview, aside from these proposals, of the problems that internet addressing has developed over the years was also presented at the same working group meeting. https://datatracker.ietf.org/doc/draft-jia-in
230.
by dtaht 5y ago
It's not "one click", but flent is the goto tool we use to probe networks for a multiplicity of problems. It's available for most OSes, including OSx. See flent.org.
231.
by dtaht 5y ago
Over here we tuned up virgin media fiber to have the least latency under normal working conditions yet seen, using sqm and the sch_cake algorithm. Pretty graphs here: https://forums.overclockers.co.uk/threads/virgin-med
232.
by dtaht 5y ago
Other normal traffic on the machine going through that conditioner has side effects on your application as well.
233.
by dtaht 5y ago
A host nowadays consists of one or more vms, piled on top of a container, piled on top of a hypervisor.
234.
by dtaht 5y ago
We have been at this, technically, for years and has been discussed in multiple forums prior to now. It's a done project for me (my primary focus is bufferbloat), as all the technical work I could do is complete and the core patches fo
235.
by dtaht 5y ago
Please contact me via email - dave dot taht at gmail dot com. Working on a successor to this: http://www.taht.net/~d/fcc_saner_software_practices.pdf
236.
by dtaht 5y ago
filed with the fcc, 6 years back: http://www.taht.net/~d/fcc_saner_software_practices.pdf
237.
by dtaht 5y ago
We could use an 'upgrade in place': https://docs.google.com/document/d/1T21on7g1MqQZoK91epUdxLYF...
238.
by dtaht 5y ago
At current prices .mil selling off 22/8 would net them at least 2 F-22s. Not quite the same as selling off tsla stock, but...
239.
by dtaht 5y ago
the zeroth draft gives folk with an allocation a free new IP address.
240.
by dtaht 5y ago
I'd like that too, having spent decades on ipv6. Making 0/8 "just work" took an hour and some testing. Ripping out the check for that reservation saved a nanosecond. And: 240/4 had been working for a decade, and nob

More ›