My attempt at building a Raspberry Pi Supercomputer

RE: My attempt at building a Raspberry Pi Supercomputer

Former Member — Fri, 08 Feb 2013 16:43:21 GMT

Just finished mine. Worked great! Documented it at http://www.rpural.net/mpich

RE: My attempt at building a Raspberry Pi Supercomputer

Former Member — Fri, 01 Feb 2013 22:35:51 GMT

Thanks for the help Ben. I will definitely have to look into this. I look forward to experiment with this.

JoelQQ

RE: My attempt at building a Raspberry Pi Supercomputer

niftymitch — Fri, 01 Feb 2013 20:00:58 GMT

I have built one of these small clusters and it is a grand way to explore parallel programming on a budget.

Step one is to get a couple two, four or five systems up running and identical.

Start with wheezy raspbian because the hard float (hf) ABI is an advantage.

Update it: sudo apt-get update; sudo apt-get upgrade

Step two is to network all the boards so you can ssh to each without a password.

Step three is to compile and load one of the current versions of MPI (message

passing interface) software. The latest compiles and runs well enough that

it is worth checking for a prebuilt version that can be installed vi apt-get or

aptitude. Doing this will require compilers and run time libraries.

Step next; compile a small standard benchmark, copy it to each of the nodes (boards)

and run it. To coordinate this the tool "mpirun" is often the correct choice however

on large clusters a queue manager is used to share, queue, bill and otherwise

manage the systems.

Results: correct, excelent teaching tool but not too speedy.

What is MPI and how does it work: One way to think about MPI is

to start with a very long tape from a ten-key adding machine or a grocery store.

By cutting the tape into N bits N people can add up their bit of tape

then those sub-answers can be merged and "finalized". The messages

passed from rank0 to other ranks can be data, offsets into a file, file name per rank,

progress and more. One analogy is to have N people sitting arround a

table. Pass mesages on 3x5 colored note cards and file folders. Compare and

contrast a round table, a long skinny rectangle table.

Important and exagerated issues common to all MPI clusters. Latency:

latency is never low enough and programming to manage latency is a bonus.

Bandwidth: bandwidth is key moving data unnecessarily, twice or more slows down the

fastest cluser. Blunders here are obvious. Algorithms: there are many ways

to skin a cat/ divide up a problem and when well chosen speed things a lot.

ETC.....

Value: Perfect for the goals of R-Pi i.e. teaching... students can build and

learn how to use clusters for pennies on the dollar. If each student has

two boards he/she can start. If a class of 16 each brings their boards to

a lab the 2 node homework can be run on 2, 4, 8, 16, 32 nodes (ranks)

evaluated, profiled, debugged. Batch and billing systems developed and tested..

Slurp on a smoothy and ponder "Slurm" without getting "Torqued" out.

Note: MPI can be coded using with C, FORTRAN, Python, C++

MPI is not the only way to run code in parallel. Parallel programming

is hard yet it is currently the best way to solve many problems.

Many multi core laptops can run circles arround a Pi cluster but

you cannot interconnect 16 student laptops together in a student lab

with the same ease and results.

SUMMARY: Ponder how one learns how to use a million+ rank (core) cluster. If each

core is about a half a grand each.... well you fill in the blanks and do the math.

The R-Pi is a win.

RE: My attempt at building a Raspberry Pi Supercomputer

Former Member — Fri, 01 Feb 2013 14:39:04 GMT

We're a group of college students at UIUC - the University of Illinois at Urbana-Champaign and we built something similar last november. We have one with 17 Pi's sitting in my room now. We found that we had much better luck using openmpi rather than mpich2. We still haven't done anything big with it yet - just some scripts to sort numbers across the distributed platform. Here's our site: http://raspi.rohankapoor.com

RE: My attempt at building a Raspberry Pi Supercomputer

rcomeau — Thu, 31 Jan 2013 23:57:03 GMT

You show this (as an email) like it's a done thing, but there is no substance to this post!!!! If you are going to make a statement about a super computer(raspberry pi version) then have something to show!!

RE: My attempt at building a Raspberry Pi Supercomputer

Former Member — Thu, 31 Jan 2013 23:22:29 GMT

So I am a little confused on how you would actually physically use this setup once you have completed all the steps provided. Would you use one Raspberry Pi like you usually would and it would automatically use the extra processing power?

JoelQQ