[Iccrg] Re: [Tmrg] Testbed Study on IW10 vs IW3

[Scheffenegger, Richard]

Hi Jerry,

Just one though regarding queue size: 

IMHO, it's probably rare to go from 1Gbps down to 64kbps. I would think
that the link from the provider side to the home user side will go
though consecutive steps of decreasing bottleneck bandwidth. Even if
some intermediate links have higher capacity again, that should in
effect alleviate the burst load on the ultimate bottleneck queue
somewhat. In such a scenario, you may get away with only 10 packets
queue depth at each step, if at least 4 consecutive smaller bottlenecks
are on the path from the sender to the receiver...

Also, Lars seems to do some investigation on home gateways:

>>> On Fri, Jul 9, 2010 at 12:15 AM, Lars Eggert
>> <lars.eggert at nokia.com<mailto:lars.eggert at nokia.com>> wrote:
>>> Hi,
>>>
>>> a quick status update. We now have received over 100
>> donated home gateways, plus a DSLAM. The students are on
>> their summer break, after which we'll start running a
>> significantly expanded set of tests over this much larger
>> population of devices.
>>>
>>> Many of yo have donated boxes and suggested more
>> experiments and better ways of performing our current tests -
>> thank you!
>>>
>>> In case you are attending IETF-78 in Maastricht and would like to
>>> donate a home gateway, simply bring it. (Or contact me now for
>>> shipping details; no cost to you.)
>>>
>>> We're especially interested in devices from outside the EU
>> and North America, or any other model we may not have yet
>> (see
>> http://fit.nokia.com/lars/tmp/2010-hgw-study-devices.txt).
>> And we're still lacking a CMTS for testing cable modems...
>>> 

Perhaps this team could provide some realistic ranges of the queue
depths (down- and upstream) for very small bottleneck pipes too?

Best regards,


Richard Scheffenegger

> -----Original Message-----
> From: Jerry Chu [mailto:hkchu at google.com] 
> Sent: Montag, 15. November 2010 07:53
> To: Stefan Hirschmann
> Cc: tmrg; iccrg at cs.ucl.ac.uk list
> Subject: [Iccrg] Re: [Tmrg] Testbed Study on IW10 vs IW3
> 
> On Sun, Nov 14, 2010 at 2:48 PM, Stefan Hirschmann 
> <krasnoj at gmx.at> wrote:
> > Jerry Chu wrote:
> >>
> >> I believe TCPM is a superset of the two lists above. If 
> not, please 
> >> find my presentation at the TCPM meeting at
> >>
> >> http://www.ietf.org/proceedings/79/slides/tcpm-0.pdf
> >>
> >> Our testbed data can be found at
> >> http://research.csc.ncsu.edu/netsrv/?q=content/iw10
> >
> > Thanks for your data. I missed information about the maximum queue 
> > size in this presentations. Can you add this informations. 
> I believe 
> > the maximum
> 
> See the fourth bullet on slide7. (Did you bother to read the 
> slides at all?)
> 
> > buffer size is for Slow-Start simulations really important.
> 
> Sure. We've tried smaller buffer size. See slides from the 
> previous IETF for the case of 20Mbps. For 64Kbps the numbers 
> turned sour quickly when max qlen goes below 40 packets. We 
> can rerun some numbers if you think it's worthwhile (and you 
> do have the time to read the results). We found out when max 
> qlen was too small not only the numbers were horrible for 
> both IW3 and IW10 but also extremely unstable.
> 
> Jerry
> 
> >
> > Stefan
> >
> 
> _______________________________________________
> Iccrg mailing list
> Iccrg at cs.ucl.ac.uk
> http://oakham.cs.ucl.ac.uk/mailman/listinfo/iccrg
>