[R-sig-hpc] Newbie question to Rmpi
Takatsugu Kobayashi
taquito2007 at gmail.com
Mon Aug 22 08:58:44 CEST 2011
Hi
I am a newbie to parallel computing, let alone Rmpi, and appreciate if you could
give me advice on R parallel computing in general.
I am a database marketing analyst and was told by my boss to do some research
to see if we could set up a parallel computing system on our current
server system
at work. This is because some of our clients asked us to do
statistical analysis on
a large data (about 30 million records * 12 columns) - multinomial
logit model and
some neural network analysis.
Our server spec is like
2* Xeon 5630
16GB Memory
2*500GB
and we are planning to purchase another set of this server for
small-sized cluster
computing. While I was reading
http://www.bioconductor.org/help/bioconductor-cloud-ami/
the following questions occurred to me and I would like to ask you:
1. Could you share websites/books/journals that I could deepen my understanding
of how MPICH2 and R work on linux?
2. With our current server system, is this code correct?
"mpirun -np 1 --hostfile /usr/local/Rmpi/hostfile R --no-save -f
/usr/local/Rmpi/xxxx.R --args 8 "
3. If I use Amazon EC2 and initiate 4 extra large instances, is this
code correct?
"mpiutil -a accessid -s keyid -w 16 -n clustername -t m1.large -v volumeid"
"mpirun -np 1 --hostfile /usr/local/Rmpi/hostfile R --no-save -f
/usr/local/Rmpi/xxxx.R --args 16 "
Sorry for my fundamental and perhaps invalid questions.
Many thanks in advance.
Taka
More information about the R-sig-hpc
mailing list