Re: OT - End-to-End QoS Network Design - Second Eddiont RANT

That's a good point, Carlos. I guess since Dave was referring to a QoS
design book, I made the assumption that the author was discussing
purposeful, calculated buffering in the final stage and not buffering in
transit. Buffering during transit will surely introduce jitter.

On Sun, Jun 8, 2014 at 5:47 AM, Carlos G Mendioroz <tron_at_huapi.ba.ar> wrote:

> John, Adam,
> as you say, I'm with Dave here (partly).
>
> I would not cross out a whole book because of a sentence. Books do have
> literary words, so to say, introducing topics and making the big
> picture. I guess this is in the making the big picture area.
> Also, we all have rights and wrongs. Nobody is perfect, we have to learn
> to assimilate the rights and weight the wrongs.
>
> To the detail, even if buffers are used in the final stage of
> transmission to compensate for jitter, this requires a synchronised
> dequeueing that is not there in the network, so I do agree that any
> buffering in transit is only there not to loose packets, at the expense
> of introducing random latency (and therefore, jitter).
>
> My 0.02.
> -Carlos
>
> Adam Booth @ 07/06/2014 21:20 -0300 dixit:
> > Hi Dave,
> >
> > I agree with John here, adding buffering can reduce interpacket delay
> > variation, particularly useful for circuit over packet and voice but
> > possibly your concern is with bufferbloat which is a different beast as
> to
> > what appears being discussed
> >
> > The point of the playout or dejitter buffer is to help build some
> > determinatistic delay for the services which are travelling on a
> > connectionless environment, allowing you to pace out the data on network
> > egress at the expected service bit rate hopefully not getting buffer
> > underruns (or in some cases overruns) - these buffers are something you
> > typically place at each end of the service termination though, not at
> each
> > network hop since they are concerned with individual services rather than
> > macroflow behaviour. These values usually made as small as practically
> > possible but some headroom is needed to handle network disruptions.
> >
> > I cant say I know the particular book you are reading but usually early
> on
> > general concepts are discussed and then get fleshed out in more detail as
> > you go through, possibly talking about areas most people can directly
> > influence first and perhaps get quick wins if there are problems and then
> > discuss more detailed behaviours and when the first introduced concepts
> > aren't necessarily a one size fits all.
> >
> > Cheers,
> > Adam
> >
> >
> >
> > On Sun, Jun 8, 2014 at 12:54 AM, John Neiberger <jneiberger_at_gmail.com>
> > wrote:
> >
> >> b Buffering does reduce jitter. It also increases latency, but has the
> >> benefit of smoothing out variable latency. It's a trade-off. In the VoIP
> >> world, that's why you have a dejitter buffer. It smooths out small
> >> variations in delay, but adds overall delay. The trick is finding the
> >> balance that solves your problem but doesn't make people feel like
> they're
> >> on a satellite link. b
> >>
> >>
> >> On Fri, Jun 6, 2014 at 1:59 PM, Dave Serra <maybeedave_at_yahoo.com>
> wrote:
> >>
> >>> B
> >>> I just cracked open the b End-to-End QoS Network Designb
> >>> Second Edition
> >>> book expecting to really hone in my understanding of QoS and
> >>> right on the
> >>> first page the author makes the following statement b streaming
> >>> video
> >>> (unbidirectional flows that benefited from both network and application
> >>> level
> >>> buffering to offset variations in transmission delays)b B We know
> >>> b variations in transmission delaysb
> >>> to be jitter.B Why is this CCIE
> >>> telling
> >>> me that network buffering reduces jitter when we know it is going to
> >>> INCREASE
> >>> jitter.B Ib m sure some willB rationalize
> >>> corner cases where jitter
> >>> can be improved by network buffering but in the
> >>> statistical random switching
> >>> world where flows come into and out of existence
> >>> the variance will
> >>> certainlyB be high.
> >>> So now Ib m left wondering if I should just put this book
> >>> down and consider its cost a donation to ciscopress?B Its only the
> first
> >>> page
> >>> and I am very
> >>> unimpressed.B
> >>> Make a small loan, Make a big difference -
> >>> Kiva.org
> >>>
> >>>
> >>> Blogs and organic groups at http://www.ccie.net
> >>>
> >>> _______________________________________________________________________
> >>> Subscription information may be found at:
> >>> http://www.groupstudy.com/list/CCIELab.html
> >>
> >>
> >> Blogs and organic groups at http://www.ccie.net
> >>
> >> _______________________________________________________________________
> >> Subscription information may be found at:
> >> http://www.groupstudy.com/list/CCIELab.html
> >
> >
> > Blogs and organic groups at http://www.ccie.net
> >
> > _______________________________________________________________________
> > Subscription information may be found at:
> > http://www.groupstudy.com/list/CCIELab.html
> >
> >
> >
> >
> >
> >
> >
>
> --
> Carlos G Mendioroz <tron_at_huapi.ba.ar> LW7 EQI Argentina

Blogs and organic groups at http://www.ccie.net
Received on Sun Jun 08 2014 - 08:38:56 ART