linking multiple -xM's a possibility?

Do the beagleboard -xM’s have the ability to be linked to each other? In other words, can I link n -xM’s together to do some number crunching? If so how would I go about doing so?

  • Josh

MPI can talk over IP. Connect the boards to an ethernet switch and
you'll have a low-power computational cluster. You could also use a
job management system, like e.g. Condor. This could turn into a very
nice research project.

j.

Forgive me if I sound a little dumb, but what is MPI? Unfortunately, a router is not an option on what I am planning.

Forgive me if I sound a little dumb, but what is MPI?

Commonly used communication protocol for parallel computation.

Unfortunately, a router is not an option on what I am planning.

Switch, not router. Tell a bit more about your plans, so maybe you
will receive a helpful answer.

j.

lol, sorry about that. Ok, my plans are to have the ability to run path finding algorithms, kalter filtering, and all data crunching on board my UAV. One of the boards will have to interface with an arduino, that runs flight controls. Since the original plan was to stream all data back to a ground system, crunch the data and then transmit course corrections back to the craft, I knew it was a huge waste of resources until I found this. Now the plans are to crunch everything on board and be able to course correct in flight with only transmitting telemetry data and video back to the ground unit and/or the android / i app.

You can connect two boards directly using a crossover ethernet cable.
For more boards you can install a small ethernet switch. To save on
weight, get rid of the enclosure and use just the bare board. Of
course all cables have to cut to length.
In this case probably you will have to very carefully think over
workload distribution and assign jobs to each board statically. You
can still use MPI for communication, but it will be probably an
overkill.

j.

2011/11/22 Josh Gooding <josh.gooding@gmail.com>

lol, sorry about that. Ok, my plans are to have the ability to run path finding algorithms, kalter filtering, and all data crunching on board my UAV. One of the boards will have to interface with an arduino, that runs flight controls. Since the original plan was to stream all data back to a ground system, crunch the data and then transmit course corrections back to the craft, I knew it was a huge waste of resources until I found this. Now the plans are to crunch everything on board and be able to course correct in flight with only transmitting telemetry data and video back to the ground unit and/or the android / i app.

are you sure your UAV can lift multiple Beagles with required batteries?

Perhaps, you should concentrate on your software to optimize it enough to run on single Beagle.