[Beowulf] GPU based cluster fo machine learning purposes

[Piotr Król]

On Thu, Apr 10, 2014 at 01:35:15PM +0000, Lux, Jim (337C) wrote:
> 
> 
> On 4/9/14 3:54 PM, "Piotr Król" <pietrushnic at gmail.com> wrote:
> 
> >3. I'm not electrical engineer but I wonder if there could be problem
> >with powering up 32/64 nodes at once. There are no wattage
> >characterization data for this board right now, but I saw some
> >informations that this board should be sub-10W.
> 
> 
> That depends more on your power supplies.  Do these boards have on board
> DC/DC converters? Or do they run off a standard PC power supply with +5,
> +12/-12, etc.
> 
> Bear in mind that 100 wall warts is a packaging challenge.
> 

Yes, board contain TI TPS51220 power controller with +5V and +3.3V
output. 12V AC power supply is delivered with Jetson DevKit.

> >
> >4. Theoretical max for this platform is 326 SP GFLOPS, I was able to
> >confirm that DP/SP ratio is 1/24 so theoretical max for DP is 13 GFLOPS.
> >Can someone elaborate or point me to documentation how hard will be to
> >utilize this power assuming CUDA and MPI usage.
> >
> >5. Operating system reside on eMMC, are there any reasons to switch to
> >SD card or SSD disk (there is a SATA port on board) ?
> 
> 
> What is your node-node interconnect fabric?  GigE and a multi port switch?
> For your task, is it going to be communication bound or compute bound?

There is GbE on board, so I assume (if I will choose this board) that
GbE switch would be best choice. I don't know exact algorithm profile
but some evidence show that it would be more compute bound.

> 
> If you don't know.. Build a small cluster, use commodity ethernet as the
> interconnect, and give it a try.
> 
> You can learn a whole lot from building a 5-10 node cluster.  Lots of
> things like packaging, interconnects, cables, etc.
> 

For sure I would like to run some microbenchmarking, unfortunately I
don't have code. I run some MPICH examples on my two old laptops and
workstation, so I have some basics, but right now trying to choose
hardware without knowing much about workload.

> Just crummy enough that I fought network problems for a couple of weeks
> (not realizing they were network problems.. Small embedded computers often
> don't have good diagnostic capability.. Hmm what can I figure out using
> busybox, only?)

New boards like Jetson or BeagleBone usually came with some modified
Debian/Ubuntu distro so planty of packages are already available.

Thank you for your comments Jim.

Regards,
Piotr Król