[net2-wg] Meeting Notes 6/1/2006

[Rodrigo Fonseca]

> In the 2000s,
> none of the participants that were part of the Emnets panel were
> participating in the standards processes and thus are knowingly
> letting bad standards emerge without academic objections.  These
> include 802.15.4b, Zigbee 1.1, and INA SP100.


Is this because the industry players make it hard for academia to join
the process, or because of a lack of interest from academia?

>
> 4) Reproducible experiments does not mean simulations--it means how do
> we reproduce experiments in the field.  SP is the first step IMHO
> since it provides a common language for protocols to run on various
> hardware configurations.
>

Although simulations are reproducible, and valuable, it is very true
that they can't be the end of the evaluation of anything. Reproducing
experiments with real radios is really hard, and writing software that
allows one to change only one component and maintain the others the
same is a step closer to the scientific method that us computer
scientists sometimes forget to follow.

Rodrigo

> -Joe
>
> On 6/1/06, Kyle Jamieson <jamieson at csail.mit.edu> wrote:
> > Attending: Sukun, Arsalan, Geoff, Kyle, Phil, Rodrigo
> >
> > Phil: anyone able to test collection significant?
> >
> > Sukun: A & I are in charge of telosb test.  Send was returning !=
> > SUCCESS, so [inaudible].  Every 1 sec each node sends pkt, timer fired
> > correctly, led toggles, Send.send returns != SUCCESS all the time.
> >
> > Phil: did you start the collection service, start the radio?
> >
> > Sukun: yes, seems they are wired through collection control
> >
> > Phil: modified app?
> >
> > S: yes, modified from TestNetwork
> >
> > P: try running in TOSSIM?
> >
> > S: telosb.  No, I should try.
> >
> > S: One q about semantics.  When root receives a packet, does it forward
> > to uart?
> >
> > P: no, signals Receive.  Depends on what the app wants to do.
> >
> > Kyle: TestNetwork forwards it.
> >
> > P: anyone else tested collection?
> >
> > [general no]
> >
> > P: will test LE.  LE seems to be okay, will test in TOSSIM this afternoon.
> >
> > R: would be great to have repository of topologies.
> >
> > P: will see if I can do something about that.
> >
> > R: able to run GDB on the python process.  can attach to that process.
> >
> > P: yes.  can put something like that in the tutorial.  especially useful
> > when you dump core.  Geoff, how is motelab looking, testing collection?
> >
> > Geoff: will start on that.
> >
> > P: I'll make sure Send works at least in TOSSIM.
> >
> > K: Sukun, how did you modify test network?
> >
> > S: removed dissemination part.  changed interval from 10 sec to one sec.
> >
> > R: I'll do TOSSIM tests on tree, and also on Mirage.
> >
> > P: I'll get started on TOSSIM asap, then testbed testing will commence.
> >   Original plan was for t2 beta rel one week from tomorrow, but now t2
> > code freeze one wk from tomorrow.
> >
> > R: what happens between code freeze & rel?
> >
> > P: test pkgs, test build/install rpm, compilation clean, docs (since
> > code is frozen)
> >
> > R: code freeze?
> >
> > P: besides bug fixes, no modifications.  reasonable deadline?
> >
> > R: yes
> >
> > S: A & I can put more time into testing on telosb
> >
> > P: if motelab is happy, you'll get a chance.  If each testing group
> > could spend one day, that would be plenty good.  Earlier the better to
> > discover problems.
> >
> > P: EmNets discussion on premature standardization.  I tossed out idea
> > that we are specifying interface to collection & wire formats, state
> > machines.  No howls & protests.  No protests => might be time, and we'll
> > have interoperability.
> >
> > K: Dave Clark was there moderating, talking about the Internet in the 70's.
> >
> > G: gives people something to compare against.
> >
> > R: evidence that this can go a long way.
> >
> > K: EmNets discussion on repeatable experiments in CS/our community.
> >
> > P: better set of TOSSIM topologies.
> >
> > K: Internet topologies.
> >
> > P: controlled set of simulation topologies.
> >
> > R: webpage/directory in the cvs tree with topologies, graphviz scripts.
> >
> > P: yes, would be neat.  Look at June 9 deadline, go back to TEPs, make
> > sure they match what the services do before this deadline.  Sukun, you
> > were interested in writing up the protocol side of things, still interested?
> >
> > S: still on my plate, but focus could be redistributed depending on
> > situation.  Which would be more beneficial for me to focus on?
> >
> > P: after code freeze.   maybe separate tep on protocols.
> >
> > S: sounds good.  agree that protocols should wait until after code freeze.
> >
> > K: let's talk about EmNets, maybe?
> >
> > P: ok, will power follow Moore's Law question.  Leakage current.  Need a
> > voltage regulator to do dynamic voltage scaling.  Power loss from regulator.
> >
> > K: really?
> >
> > P: DVS is operating in laptops/desktops.  First chips were 200-400 MhZ
> > arm processors.  Power tradeoffs very different.
> >
> > K: really not proportional?
> >
> > P: not clear overhead scales.  mica platform had voltage regulator.  had
> > to remove voltage regulator for low power operation.  interesting
> > questions...
> >
> > K: Columbia paper on low power compilation.
> >
> > P: Summarizes paper, 10-15% variance in energy.  If you issue two
> > instructions rather than one, can save energy.  Load indirect vs. load
> > absolute.
> >
> > P: 4.3% decrease in energy consumption.
> > _______________________________________________
> > net2-wg mailing list
> > net2-wg at millennium.berkeley.edu
> > https://mail.millennium.berkeley.edu/cgi-bin/mailman/listinfo/net2-wg
> >
> _______________________________________________
> net2-wg mailing list
> net2-wg at millennium.berkeley.edu
> https://mail.millennium.berkeley.edu/cgi-bin/mailman/listinfo/net2-wg
>