[Cu-wireless] pieces of the rooftop networking puzzle?

[David Young]

All,

Here are some ideas in wireless networking that I think belong in a
"roadmap" for development of grassroots rooftop networking. Most of
these ideas come from research papers or from the commercial sector.

KarlNet's (karlnet.com) TurboCell uses packet aggregation and a polling
protocol to improve performance in 802.11 networks. I think that
techniques such as those will be important for us to try if RTS/CTS
cannot solve some performance problems we are having right now.

Even if it turns out that Hazy-Sighted Link-State routing
is not an easy-to-implement, scalable ad hoc routing
algorithm that supports dynamic wireless links, the analysis in
<http://camars.kaist.ac.kr/~kwshin/Papers/mobihoc01v4.pdf> seems to be
an important stepping stone toward algorithms that are.

I have read a Ph.D. thesis which describes how to choose paths in a
packet-radio network which reduce the sum interference that a packet will
cause for links on the network, in order to facilitate channel re-use.
The author considered military spread-spectrum systems, but the general
idea of the thesis is applicable to multihop 802.11 networks. If
anybody would like to read the thesis, I will make one or two copies
for circulation in C-U.

Consecutive packets in the same network flow can interfere with each
other, causing delays and dropped packets. Perhaps routes should be
chosen to reduce "intra-flow interference." Also, my shallow analysis
shows that by adding a second radio and omni antenna to each node, you
can reduce intra-flow interference by enough to triple TCP performance.
Or am I on crack?

Transmit power control in 802.11 networks seems desirable not only for
interference mitigation and channel re-use, but for ensuring "fair"
medium access. It is possible for louder stations to get the attention
of a peer or access point more reliably than more quiet stations.  It is
possible with open-source drivers for the ADMtek ADM8211 (by yours truly)
and for the Atheros 802.11a/b/g chips (by Sam Leffler) to actually do
power control.

Dave

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