[Cu-wireless] HSLS?

[David Young]

On Tue, Mar 09, 2004 at 08:43:53PM -0800, Todd Boyle wrote:
> At 07:43 PM 3/9/2004, David Young wrote:
> >  You don't really think that we can answer this question, do you? =)
> 
> 
> Thanks, no of course not, not with any precision!
> 
> But what would be your guess, based on the experience
> you've had with the mesh protocols and the senao/soekris
> hardware?
> 
> What happens when 3 or 4 nodes on the network
> simultaneously try to move big files on this network?
> http://www.cuwireless.net/images/wirelessmap.jpg
> Not asking for the worst case scenario but really,
> the scenario you'd almost expect, from bandwidth
> hungry users in the evening time..
> Would the throughput
> 1. crash to zero?
> 2.  scale down gracefully?
> 3. become sporadic, randomly, i.e. this is what
> my instincts tell me, but that's based on Internet
> and office LANs when they get overloaded..
> there would be unpredictable latency
> 
> Just looking for your guess, ol boy.  You know your
> gear by now... overall, do we get 100Kbps in this situation,
> or 500kbps or 1Mbps?

My best guess is that it will surprise us.  I think that you are aware
as I am that there is research showing instability of multiple TCP
flows on 802.11.  IIRC, it has to do with unfairness that arises from
contention window scaling.  The first station to enter the medium has an
advantage the next time.  New radios seem to let us diddle with the CW,
or else to examine it.  Maybe we can take good advantage if we observe
TCP instability.

BTW, if somebody will produce a plan to run and interpret bandwidth trials
on the existing network, I will do all in my power to help you run them.

Dave

-- 
David Young             OJC Technologies
dyoung at ojctech.com      Urbana, IL * (217) 278-3933