[hen-mgmt] Re: [Sensors] Mote testbeds

[Adam Greenhalgh]

PoE motes do exists, and they don't add that much to the cost, but a
cost benefit analysis is something we need here. The big thing we have
to remember when setting this stuff up is that we have next to zero
support allocated to hen, so we should automate as much of the control
of the sensors as possible and mimimise the number of components which
the sensors depend on. I am not suggesting we be very restrictive
about what we allow people to do, all I am suggesting is that the base
communications channel and power are designed in such away as they are
always avaiable for control purposes.

I would advise against using people's desktops as part of a long term
testbed as this means that kernels and packages have to be maintained.
My big conern with things that if we don't the infrastructure in a way
that is maintainable O(const) then we will end up with a pile of
hardware. This is why i am suggesting PoE. The worse thing is to try
and do an experiment and find that half of the equipment doesn't work,
it would be like someone taking the pipe off a fume cupboard.

Adam

On 19/09/05, Piers O'Hanlon <p.ohanlon at cs.ucl.ac.uk> wrote:
> Yes it had occurred to me that USB might be simpler as there appear to more
> devices that are ship with USB (though I'm no expert). Though Adam's point
> does make things tricky....however there are a couple of potential options:
>  - USB over IP hardware - This seems to be thin on the ground and currently
> a bit expensive (e.g. Digi's AnywhereUSB at around $250 for a 5 port USB
> over IP solution).
>  - Distributed USB over IP using PCs - There's quite a nice project that
> provides for extending USB over IP from one host PC (linux) to another:
> http://usbip.naist.jp
>  
>  Neither is ideal, though the second could be neat  - maybe we should be
> thinking of semi-managed machines whereby such hardware resources may be
> shared between machines.
>  
>  If there are some suitable PoE motes then I guess these would be best.
>  
>  Piers
> 
> 
> On 19/09/05, Adam Greenhalgh <a.greenhalgh at cs.ucl.ac.uk> wrote:
> > 
> > There is a downside as USB requires PCs to be near by or cabling to
> > work over a largish distances . If you look around the 7th floor there
> > is no central control of the PCs, each is administered by a different
> > person. The only common infrastructure we have that is likely to be 
> > able to scale long term is ethernet.
> > 
> > Adam
> > 
> > On 18/09/05, Brad Karp <bkarp+ at cs.cmu.edu> wrote:
> > > Ramesh Govindan at USC has told me that his lab has had great success 
> > > using USB for the mote control channel. The USB interface hardware for
> > > each mote is far cheaper than that for Ethernet, and Ramesh reports
> > > that the USB interface offers equivalent functionality.
> > >
> > > I don't recall right now if his lab is *powering* the motes using USB,
> > > as well; I'll ask him.
> > >
> > > I think we should definitely look into this USB control channel
> > > approach, as it will let us stretch our funds further, and there 
> > > doesn't appear to be any down side.
> > >
> > > -Brad, bkarp at cs.cmu.edu
> > >
> > > On Sun, Sep 18, 2005 at 08:48:33PM +0100, Adam Greenhalgh wrote:
> > > > Interestingly it would seem that all test beds in the presenstations 
> > > > linked from here use ethernet as the control channel. They also seem
> > > > to use a separate power controller. I don't think we can do this with
> > > > the setup in the current building, we have no ceiling voids to put 
> > > > things like power in. I think PoE might have to be our answer.
> > > >
> > > > Looking at PoE solutions I think we probably are going to need a mix
> > > > of a large 24 port PoE switch and some smaller desk mounted PoE 
> > > > switches to give us flexibility.
> > > >
> > > > Adam
> > > >
> > > > On 18/09/05, Dima Diall <dimadiall at gmail.com> wrote:
> > > > > Hello all: 
> > > > >
> > > > > I had this unfinished message in my drafts since August due to lack
> of
> > > > > time near the end of the final project. It was intended to be a
> > > > > follow-up to the (first?) "UCL-CS Motes Meeting", sponsored by Steve
> > > > > Hailes (in late July)... I'll just finish it off with what I still
> > > > > remember, regarding the department's plan to set up a mote testbed.
> > > > >
> > > > > I think the first approach was the "ceiling array" at Berkeley: 
> > > > >
> https://mail.millennium.berkeley.edu/pipermail/tinyos-help/2005-July/010708.html
> > > > >
> > > > > Furthermore, in the last TinyOS Technology Exchange Mote four
> projects 
> > > > > related to testbed deployment and management were presented. Links
> are
> > > > > provided on the TTX II page (search under the heading "Embedded
> > > > > Wireless Networking Testbeds"): 
> > > > > http://www.tinyos.net/ttx-02-2005/
> > > > >
> > > > > There also is a working group on TinyOS testbeds but with very
> little
> > > > > activity so far: 
> > > > >
> http://www.tinyos.net/scoop/special/working_group_tinyos_testbed
> > > > >
> > > > > At the time when I started this I had more details, but it has 
> > > > > vanished by now ;-)
> > > > > --
> > > > > /dima, http://www.cs.ucl.ac.uk/students/d.diall/
> > > > >
> > > > > _______________________________________________ 
> > > > > Sensors mailing list
> > > > > Sensors at cs.ucl.ac.uk
> > > > > http://oakham.cs.ucl.ac.uk/mailman/listinfo/sensors
> > > > >
> > > >
> > > > _______________________________________________
> > > > Sensors mailing list
> > > > Sensors at cs.ucl.ac.uk
> > > > http://oakham.cs.ucl.ac.uk/mailman/listinfo/sensors
> > > >
> > >
> > 
> > _______________________________________________
> > hen-mgmt mailing list
> > hen-mgmt at cs.ucl.ac.uk 
> > http://oakham.cs.ucl.ac.uk/mailman/listinfo/hen-mgmt
> > 
> 
>