Skip to content

Parallel R

5 messages · Bazman76, R. Michael Weylandt, Whit Armstrong

#
Hi there,

I have a largish optimisation problem (10 years of daily observations).

I want to optimise between 4 and 6 parameters.

I'd like to utilise parallel computing if I can as I will have to run it
with different starting values etc.

I have a quad core PC with 16GB ram running windows 7.

Having done  a little reading it seems the two best options for me are:

1.) Use the academic version of REvolution R enterprise which will let me
run the code in parallel on my own PC.

2.) Or install Unubtu and use Segue and Amazon EC2 (or some equivalent).

Just to get an idea how much faster is option 2 liable to be (given that its
slightly more difficult to set up). 

Is it realistic to have a dual boot on my PC?

Thanks

Baz



--
View this message in context: http://r.789695.n4.nabble.com/Parallel-R-tp4643175.html
Sent from the R help mailing list archive at Nabble.com.
#
On Fri, Sep 14, 2012 at 6:00 PM, Bazman76 <h_a_patience at hotmail.com> wrote:
"Regular" R allows for parallel computation also on all supported
platforms (including Windows)... look at the included 'parallel'
package. Some optimizers (DEoptim perhaps?) will have
auto-parallelization available as well.

Cheers,
Michael
#
Thanks for that I hadn't realised parallel could run on windows Pc's.

The code is differential evolution but its not part of a package.

Still I would like to be able to use cloud computing if possible, any
thoughts on the easiest way to achieve that using a windows based PC?

Found this blog which seems to give pretty clear instructions:

http://travisnelson.net/2011/05/05/ec2-micro-instance-of-rstudio/

and was also recommended cloudRmpi package.

Have you experience with either of these?

 





--
View this message in context: http://r.789695.n4.nabble.com/Parallel-R-tp4643175p4643186.html
Sent from the R help mailing list archive at Nabble.com.
#
On Fri, Sep 14, 2012 at 7:22 PM, Bazman76 <h_a_patience at hotmail.com> wrote:
Why does your local operating system dictate the backend OS used by
the cloud servers? I'd use what your servers already have on them -- R
should run more or less identically. (For the record, I believe linux
backends will be more performant in general)
Please quote context for the (vast majority of) readers who don't use Nabble.

No, I don't have experience with those, but you may try the R-SIG-HPC
list for a more focused audience.

Cheers,
Michael
#
I addition to Michael's suggestions, you can also check out this
tutorial which shows how to use lapply into EC2.

http://www.rinfinance.com/agenda/2012/workshop/WhitArmstrong.pdf

Unfortunately, rzmq is not available on windows, so this may not be
the best solution for your setup.

-Whit
On Fri, Sep 14, 2012 at 2:22 PM, Bazman76 <h_a_patience at hotmail.com> wrote: