[Cu-wireless] C-U Wireless Status Update

[David Young]

An update on the wireless project is long overdue.

Grants
------

    C-U Wireless won a $4500 grant to buy tools and rooftop routers.
    We have put together a kit for installing rooftop routers, and we
    have also bought the parts to construct six routers. Each router
    resembles the ammo can router we use at IMC: it consists of a Soekris
    single-board computer, a weatherproof enclosure, 8dBi antenna, an
    802.11 radio, CompactFlash card, 50 feet of outdoors-proof ethernet,
    and a PoE injector. We will use these nodes to spread the network
    northeast from Zach Miller's house, just as soon as our order for
    CompactFlash cards is (re-)fulfilled with the right part.

    We have submitted a grant proposal to George Soros' Open Society
    Institute. I think our chances of getting this grant are pretty slim,
    however, the time spent writing the grant was well spent because
    I think we have strongly articulated the project philosophy and
    goals. Also, the grant contains a comprehensive technical development
    plan, time and cost projections, etc. Expect to see excerpts from
    the text on the mailing list and on the Web.

Software Improvements
-------- ------------

    Here and there I have made an improvement to the software
    distribution.  It fits in just 32MB, now, so we can fit an
    "experimental" and "last good" copy onto one 64MB CF card. Trouble
    is, we cannot revert to the "last good" copy if the "experimental"
    version fails. If somebody wants to port the FreeBSD bootloader
    or else to hack the NetBSD bootloader to give us a "boot once"
    capability, it would be an ENORMOUS help. I just don't have the time
    to do all this stuff.

    I have ported from FreeBSD to NetBSD the driver for 802.11a/b/g cards
    based on the Atheros chipset.  Great improvements to the 802.11 layer
    accompanied that driver.  (NetBSD & FreeBSD contain a *nice* 802.11
    networking layer that makes it easy to provide uniform support for
    lots of 802.11 gear, especially the new 802.11 ASICs. Because of the
    802.11 layer, I think that you're going to see NetBSD leapfrog Linux
    and OpenBSD in terms of open-source drivers for new 802.11 radios.)

    I don't think that we can take good advantage of a/g modulation
    (needs too-high S/N ratio), but Atheros-based 802.11b cards are a
    LOT more versatile than Prism-based cards. I want to eliminate Prism
    cards from the C-U Wireless network ASAP ... they just don't expose
    provide enough knobs and dials for one to do interesting things on
    rooftop networks.

    I've programmed an uncomplicated rate adaptation algorithm that I
    read about in someone's research paper. It is more sophisticated
    than Lucent's algorithm for choosing the transmission rate (1, 2,
    5.5, 11Mbps), which is what most people use. A decent rate adaptation
    algorithm will improve performance on the C-U network.

    I tried and failed to make the CUW software distribution run both a
    DHCP client and server on each interface, so that any port can act
    as a WAN or LAN port, so that you can build a true "plug & play"
    network. I think that I will try again a little later.

New Directions for Rooftop Routing
--- ---------- --- ------- -------

    I've mentioned before that the MIT Roofnet's ETX routing metric
    (or something like it) is something we must start using before
    we have much greater density of nodes, or else the shortcomings
    of shortest-path routing on wireless are going to show up with
    a vengeance.

    My software development plan describes the architecture for a routing
    daemon for rooftop networks that uses the ETX routing metric.

    There is a Ph.D. thesis (written in 1980!) that has really influenced
    the way I think about wireless networks. I mean, as if it wasn't
    clear enough, routing in wireless networks is a LOT different than
    routing in wire networks, mainly because of (self-)interference.
    The Ph.D.  thesis is concerned with a link-state routing protocol
    called "Least Interference Routing." I have made two spiral-bound
    copies of the thesis if you want to find out more.

Developers Needed
---------- ------

    If you are an experienced C programmer, I could use your help
    to develop the next generation of routing software and also
    a pretty slick name service, 802.11 innovations, etc., for C-U
    Wireless. Network-programming experience is a plus; kernel programming
    experience is a plus-plus. Let me know if you are interested, and
    I will send you a copy of the development plan.

Research Guinea Pigs
-------- ------ ----

    Professor Christian Sandvig in Communications is interested in
    the history and potential social impacts of community wireless
    networks. Students of his will be interviewing Sascha and I as part
    of a research project on community wireless networking.

Major Interference Source Identified
----- ------------ ------ ----------

    I have found out the operator of the powerful wireless interference
    source. We are working together to solve the problem. It turns
    out that the interference comes from channel 6 (we are on channel
    11). It is possible that a defective radio (no filtering on our end)
    or a software defect (their end) is to blame; it is also possible
    that our systems are just way too close to each other to co-exist
    on channels 6 and 11.

Hardware Explorations
-------- ------------

    I don't think any outdoor 802.11 network can scale very big without
    using both power control and "adaptive" or "smart" antennas---that is,
    antennas that let you either choose propagation pattern presets or
    form beams in the direction of network peers. Power control helps
    you re-use the same frequency in a region (the analogy is that many
    more people can communicate in one room by whispering one to another
    than by shouting). Adaptive/smart antennas let you direct your
    radio energy directly at your communication partner: it gives you
    "spatial re-use" like power control, but at a finer level; it also
    helps you protect against interference sources.

    New radio chips (ADMtek, Atheros, Realtek) let you control power
    levels. Atheros lets you control power for each packet---very nice!
    I am trying to find out from Prof. Jennifer Bernhard if her 2.4GHz
    adaptive antenna for space communications is suitable for rooftop
    routers.

    There are already some commercial projects with smart antennas, but
    they are super-expensive. Somebody called me the other day to tell
    me about their Vivato "wireless switch" which cost just more than
    $10,000. If anybody is interested in seeing a demonstration of the
    Vivato switch, the Ohio company that called me offers to demonstrate
    for C-U Wireless next time they are in the area.

    In my naivete, I do not see any reason that a computer-controllable,
    2.4GHz adaptive antenna (less versatile and less complicated than
    a "smart" antenna such as Vivato's) cannot be manufactured on
    printed-circuit board for $150 or less.

Senior Project Team
------ ------- ----

    This year I am working with a second senior project group. The
    project for last year (roaming in 802.11 networks) involved kernel
    programming and other low-level stuff, so it was really too much
    for time-strapped seniors to finish in one year. This year, four of
    Ralph Johnson's students are working on a Java app for visualizing
    802.11 packet traces, such as tcpdump and ethereal create. The reason
    for programming a visualizer is that you can use it to teach about
    the 802.11 MAC protocol, to verify 802.11 implementations, and to
    identify/characterize interference. If you've seen the protocol
    diagrams in the 802.11 protocol spec, then you know roughly what
    the visualizer will produce.

Dave

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