Re: [Codel] some backing scripts from Multimedia-unfriendly TCP Congestion Control and Home Gateway Queue Managemen

[Kathleen Nichols <nichols@pollere.com>]

Thanks, Dave. I will try to get to this eventually...

I did some runs with 1G bw and 5ms RTT. I haven't had time
to process them, but they are okay.

On 5/1/12 8:56 AM, Dave Taht wrote:
> ---------- Forwarded message ----------
> From: Andreas Petlund <apetlund@simula.no>
> Date: Tue, Apr 24, 2012 at 11:57 PM
> Subject: Re: Bufferbloat experiment preliminary result
> To: David Täht <dave.taht@gmail.com>
> 
> 
> Hi Dave,
> 
>  The key ns2 scripts and analysis tools that was
> used in "Multimedia-unfriendly TCP Congestion Control and Home Gateway
> Queue Management" can be found here:
> http://caia.swin.edu.au/urp/newtcp/tools/cbr-drop-scripts.tar.gz
> 
> To successfully run this simulation, ns2 needs to be patched with the
> CAIA CBR traffic extension found here:
> http://caia.swin.edu.au/urp/newtcp/tools/ns2_33_patch0.1.1.tar.gz
> 
> I hope this will aid you in your work. Also hope to speak to you soon.
> 
> 
> Cheers,
> Andreas
> 
> 
> On 11/02/2011 07:46 PM, David Täht wrote:
>> On 11/01/2011 03:03 PM, Andreas Petlund wrote:
>>> On 10/31/2011 08:52 PM, Dave Taht wrote:
>>>> Working backwards...
>>>>
>>>> 1) where is "there"? I'm in paris, don't have much budget left for
>>>> travel this year.
>>>>
>>> With the new kid in town, I'll not be travelling much the first couple
>>> of months, but the next IETF meeting is in Paris, I believe. That may be
>>> a golden opportunity. I'm also available on Skype (andreas_petlund).
>>
>> Taipai (can't spell it) November 15th. Paris would be a bit of a detour!
>>
>> am davetaht on skype but rarely am on - too many distractions.
>>
>> Perhaps as we pull a plan togther, post baby, we can do a couple
>> concalls.
>>
>>>
>>>> 2) Do you mind if I put these scripts in a public place (github?) I am
>>>> slowly assembling bits from the 4+ experiements I want to run and
>>>> re-run and re-run....
>>>>
>>> I have sent an email to the other authors (the Swinburn people). If they
>>> agree (which I do believe they will), I'll give you green light for
>>> making it public.
>> OK. I also have a bittorrent and tcp-lp simulation to play with. I'd
>> like to get a decent red and qfq sim up and then hook it all up
>> together. Might be able to steal some time and vms from sandia.
>>>> 3) Which paper was this relevant to? The media-unfriendly one?
>>>>
>>> Correct.
>>>
>>>> 4) I think your work is spot-on-relevant, too, which is why we're
>>>> talking. :) In my case, I tumbled into this by accident - I was
>>>> working on mesh networks in Nicaragua... and upgraded everything from
>>>> 'g' to 'n' and thought it would be 'better'. It wasn't....
>>>>
>>> We started out with the Norwegian MMORPG game company Funcom and some
>>> traces that showed some extremely high application-layer latency.
>>
>> While it certainly exists there, it starts at the edge and works in...
>>
>> It bothers me most that nobody grokked that while tcp scales well to
>> lunar orbit, at such high levels of buffering it is very human
>> unfriendly. 100 vs 1000 buffers is quadratic in terms of human
>> friendlyness of tcp... 10,000 worse, not 1000 times...
>>
>> I need a new word to use instead of 'fair' or 'unfair' when it comes
>> to queuing - 'friendly queuing' is working for me...
>>> I've
>>> mostly been working with end-to-end transport, but have been looking
>>> into buffers recently. A lot of scenarios to explore where things that
>>> is seemingly benign really is not.
>>
>> Want a brain dump from van? I have been reading his mid-late 80s stuff
>> over and over and it wasn't until I got it straight from the source
>> before I truly got it - and even then i only get 85% of it, and
>> getting 86% would probably take 10 years of study and and a partial
>> brain transplant.
>>
>>> Cheers,
>>> Andreas
>>
>>
> 
> 
>