Re: [Bloat] Usage Based Billing - It's All About Perceived Congestion

[Jim Gettys <James.Gettys@alcatel-lucent.com>]

On 03/01/2011 08:57 AM, Steve Bauer wrote:
> On Mon, Feb 28, 2011 at 4:56 PM, richard<richard@pacdat.net>  wrote:
>> The result is:
>> http://digital-rag.com/article.php/All-About-Perceived-Congestion-UBB
>> It's aimed at the public and non-technical so allow me a few missing
>> things in the explanations :)
>
> Hi Richard,
>
> You recommended to your readers:
>
> "You want to bug your ISP to turn off any/all "helper" buffering in
> your modem that purports to "increase upload speed for short periods"
> since this is buffering and part of the problem."
>
> I am assuming you intended that to apply to Powerboost.  Based upon my
> understanding of various implementations of Powerboost,
> turning it off would *not* help with latency under load simply
> because the buffer sizes are fixed in today's cable modems.
>
> So these issues are orthogonal.  There are certainly
> problems, but turning off Powerboost isn't part of the fix.
>
> Indeed, turning Powerboost off could make the problem worse. A
> buffer that would otherwise have drained if Powerboost was in effect,
> would be more likely to have a queue of packets sitting in it adding
> additional latency to later arrivals.  Again, my key assumption is
> that disabling Powerboost simply doesn't change the buffer size.
>
> As always, happy to be proved wrong.  :-)
>

I think you are correct, Steve, and turning it off will just make the 
situation worse.

At the moment, I think these buffers are static in these technologies; 
turning off Powerboost or equivalent features will just make the 
situation worse.

We're going to have to be adaptable to variable bandwidth anyway.  AQM 
is ultimately the only solution.
			- Jim