Re: [Codel] [aqm] Codel's count variable and re-entering dropping state at small time intervals

[Dave Taht <dave.taht@gmail.com>]

this is not a day to discuss codel. it is a day to fix the internet.

http://www.businesswire.com/news/home/20151014005564/en/Global-Internet-Experts-Reveal-Plan-Secure-Reliable

Please expect any codel related discussion to go by the wayside for a while.

On Wed, Oct 14, 2015 at 9:01 PM, Agarwal, Anil <Anil.Agarwal@viasat.com> wrote:
> Toke,
>
> Here is the info you requested.
> Please let me know if you find any errors.
>
> For the example in the draft rfc,
> based on analytical equations for hash collision probabilities,
>         the probability of no collision = 90.78%,
>         probability that no more than two of the 100 VoIP sessions will be involved in any given collision = 99.57%,
>         probability that no more than three of the 100 VoIP sessions will be involved in any given collision = 99.99%.
>
> With a 4-way set associative hash table of equivalent size,
>         the probability of no collision = ~100%,
>
> With an 8-way set associative hash table of equivalent size,
>         the probability of no collision = 99.93%,
>
> Regards,
> Anil
>
> -----Original Message-----
> From: Toke Høiland-Jørgensen [mailto:toke@toke.dk]
> Sent: Wednesday, October 14, 2015 1:58 PM
> To: Agarwal, Anil
> Subject: RE: [aqm] Codel's count variable and re-entering dropping state at small time intervals
>
> Yes, my understanding is #2 (Paul wrote the section originally, will check with him). Wasn't planning to mention set-associative hashes in this draft, but we do use it in cake (the "successor"), so would love to see the analysis!
>
> Incidentally, are you aware of any other fairness queuing systems that use set-associative hashes?
>
> -Toke
>
> On 14 October 2015 19:23:51 CEST, "Agarwal, Anil" <Anil.Agarwal@viasat.com> wrote:
>>Toke,
>>
>>Also, do you plan to add some notes on the use of set-associative hash
>>tables?
>>My analysis also shows collision probabilities for N-way
>>set-associative hash tables.
>>
>>Anil
>>
>>From: Agarwal, Anil
>>Sent: Wednesday, October 14, 2015 1:13 PM
>>To: 'Toke Høiland-Jørgensen'
>>Subject: RE: [aqm] Codel's count variable and re-entering dropping
>>state at small time intervals
>>
>>
>>Toke,
>>
>>
>>
>>I want to make sure I precisely understand the following statement in
>>the draft rfc -
>>
>>
>>
>>There is about an 86% probability that no more than two of the 100 VoIP
>>
>>sessions will be involved in any given collision.
>>
>>
>>
>>Is it saying that -
>>
>>1.       There is an 86% probability that 98 sessions will be
>>collision-free and only 2 sessions will collide.
>>
>>I think not.
>>
>>or
>>
>>2.       There is an 86% probability that among the sessions that
>>collide, the number of sessions that collide together at a common bin
>>value will be 2.
>>
>>There is a 14% chance that 3 or more sessions will collide into the
>>same bin (and hence share the same queue).
>>
>>or
>>
>>3.       Something else
>>
>>
>>
>>I assume it is #2.
>>
>>Please confirm and then I will send you the analysis.
>>
>>
>>
>>Thanks,
>>
>>Anil
>>
>>
>>
>>
>>
>>-----Original Message-----
>>From: Toke Høiland-Jørgensen [mailto:toke@toke.dk]
>>Sent: Wednesday, October 14, 2015 9:13 AM
>>To: Agarwal, Anil
>>Subject: Re: [aqm] Codel's count variable and re-entering dropping
>>state at small time intervals
>>
>>
>>
>>"Agarwal, Anil"
>><Anil.Agarwal@viasat.com<mailto:Anil.Agarwal@viasat.com>> writes:
>>
>>
>>
>>> I will clean up the equations and send you the doc in a few hours.
>>
>>
>>
>>Cool, thanks!
>>
>>
>>
>>> Are in the process of revising draft-ietf-aqm-fq-codel-01? It's
>>
>>> expiration date is shown as January 5, 2016. But I suppose it is
>>
>>> useful to post the next update earlier than that.
>>
>>
>>
>>Yeah, figured I'd do another revision before the next meeting, even if
>>the AQM group is not scheduled. Current draft here:
>>
>>https://urldefense.proofpoint.com/v2/url?u=https-3A__kau.toke.dk_ietf_d
>>raft-2Dietf-2Daqm-2Dfq-2Dcodel-2D02.html&d=BQIBAg&c=jcv3orpCsv7C4ly8-ub
>>Dob57ycZ4jvhoYZNDBA06fPk&r=FyvaklKYrHaSCPjbBTdviWIW9uSbnxdNSheSGz1Jvq4&
>>m=2GbxJwOxywTH1qXvOYaEDXpi8xnj4PMSw3vn7aDEszU&s=Sj61jStdaiIE503n8_vQ9Bf
>>k-F9uk_hdcDi9x47m66U&e=
>>
>>
>>
>>-Toke



-- 
Dave Täht
Do you want faster, better, wifi? https://www.patreon.com/dtaht