Re: [aqm] PI2 presentation in iccrg

["De Schepper, Koen (Nokia - BE)" <[e-mail]>]

As there was no time for questions, I wanted to share a question/remark on the list that I got from multiple people:
"Won't the think twice with 2 random numbers at the end step be difficult to implement in existing hardware?"

This is true, but most likely the PIE algorithm is implemented in an off-line SW process and in that case the squared probability can be simply implemented by multiplication to itself, before applying it to the hardware.

The 2 random numbers are a very efficient way to implement in new hardware, because you can use half the resolution for a classic probability comparison. For example when a probability up to p = [0.000001 .. 1] needs to be supported, the p' range needs to be only [0.001 .. 1]. You do need 2 random numbers, but only half their size, so the same amount of random bits.

If there are other questions, let me know or post on the lists.

Koen.

The link to the slides: https://www.ietf.org/proceedings/95/slides/slides-95-iccrg-3.pdf

> -----Original Message-----
> From: aqm [mailto:[e-mail]] On Behalf Of EXT De Schepper,
> Koen (Nokia - BE)
> Sent: vrijdag 1 april 2016 12:32
> To: EXT Wesley Eddy; [e-mail]
> Subject: [aqm] PI2 presentation in iccrg
> 
> Hi all,
> 
> Just to inform you that I will do a presentation on PI2 in the iccrg
> session on Monday.
> 
> The focus will be on support for scalable TCPs and throughput fairness
> between classic (CUBIC) and scalable (DCTCP) congestion controls, but
> the basics of the PI2 AQM will be explained as well. It is build on PIE
> (actually a simplification) which supports also scalable TCPs and needs
> less (internal) parameters. As mentioned before this could be
> standardized as extensions on the PIE draft, or a new PI2 draft could be
> made for this. To be discussed afterwards.
> 
> Regards,
> Koen.
> 
> 
> 
> > -----Original Message-----
> > From: aqm [mailto:[e-mail]] On Behalf Of EXT Wesley Eddy
> > Sent: vrijdag 26 februari 2016 23:05
> > To: De Schepper, Koen (Nokia - BE); [e-mail]
> > Subject: Re: [aqm] AQM plans
> >
> > Hi, I'm now sorry that we didn't schedule an AQM meeting slot in
> Buenos
> > Aires.  We can certainly pre-load an agenda for Berlin with these
> topics
> > though.  That is a long way in the future, however, so please use the
> > mailing list to discuss tests and metrics, share results, advertise
> > papers, etc, in the meantime on these topics!
> >
> >
> >
> > On 2/26/2016 6:23 AM, De Schepper, Koen (Nokia - BE) wrote:
> > > Hi Wes,
> > >
> > > Just to let you know that we are still working on AQMs that support
> > scalable (L4S) TCPs.
> > > We could present some of our latest results (if there will be a
> > meeting in Buenos Aires, otherwise in Berlin?)
> > >
> > > * Performance of HTTP Adaptive Video Streaming (HAS) with different
> > TCP's and AQMs
> > >     o HAS is currently ~30% of Internet traffic, but no AQM testing
> so
> > far has included it
> > >     o the results are very poor with a particular popular AQM
> > > Presenter: Inton Tsang
> > > Duration: 10mins
> > > Draft: Comparative testing of draft-ietf-aqm-pie-01, draft-ietf-aqm-
> > fq-codel-04, draft-briscoe-aqm-dualq-coupled
> > >
> > > For experiment write-up, see Section 3 of
> > https://riteproject.files.wordpress.com/2015/12/rite-deliverable-3-3-
> > public1.pdf
> > >
> > > * PI^2: PI simplified with a square
> > >     o PIE embeds some auto-tuning and heuristics, which can be
> removed
> > by simple transformation of a variable
> > >     o Allows PIE to control also scalable congestion controls
> (DCTCP,
> > ...)
> > > Presenter: Koen De Schepper
> > > Duration: 15mins
> > > Draft: (probably impacted)  draft-ietf-aqm-pie
> > >
> > > * DualQ update
> > >     o extensive additional tests, including different RTTs and mixed
> > RTTs, better visualization
> > >     o using other AQMs than Curvy-RED for 'Classic' traffic, e.g.
> PI^2
> > > Presenter: Koen De Schepper
> > > Duration: 20min
> > > Draft: draft-briscoe-aqm-dualq-coupled-01
> > >
> > > Regards,
> > > Koen.
> > >
> > >
> > >> -----Original Message-----
> > >> From: aqm [mailto:[e-mail]] On Behalf Of Wesley Eddy
> > >> Sent: woensdag 10 februari 2016 20:27
> > >> To: [e-mail]
> > >> Subject: [aqm] AQM plans
> > >>
> > >> Hello AQMers, this is just a quick note to be clear on working
> group
> > >> status and forward planning.
> > >>
> > >> Currently, all of the active drafts are either in WGLC, or in the
> > >> process of shepherd writeups to go the the AD.
> > >>
> > >> Once we get the current set of drafts out for publication, there
> are
> > a
> > >> few things the WG can do:
> > >> - close down
> > >> - remain open to continue coordinating and advance algorithms (e.g.
> > >> within the standards track)
> > >> - remain open as a venue for potential new work (e.g. the DualQ
> > Coupled
> > >> AQM)
> > >>
> > >> I think we should discuss this and plan to make a decision after
> the
> > >> IETF 96 meeting in Berlin (July 17-22).  We can have a "closing or
> > >> re-chartering" meeting in Berlin if it will benefit from face-to-
> face
> > >> discussion.
> > >>
> > >>
> > >> _______________________________________________
> > >> aqm mailing list
> > >> [e-mail]
> > >> https://www.ietf.org/mailman/listinfo/aqm
> > >
> >
> > _______________________________________________
> > aqm mailing list
> > [e-mail]
> > https://www.ietf.org/mailman/listinfo/aqm
> 
> _______________________________________________
> aqm mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/aqm