I have a question about building R (2.15.1) from source on a Solaris 10
64bit virtual server with 2 cores and 16GB memory that is running on an
Oracle T4 server.  Based on several tests I have done this configuration
has been several orders of magnitude slower than any other configuration
I've tested.  

A simple test of some code to calculate the standard deviation 10000
times (simple code to consume resources) takes on average 121.498
seconds on the Solaris server where the next worst system (Redhat Linux)
takes 1.567 seconds:
t0 <- proc.time()[[3]]
x <- rnorm(10000)
for(i in 1:10000){
sd(x)
}
print(proc.time()[[3]]-t0)

R-Benchmark (http://r.research.att.com/benchmarks/R-benchmark-25.R) is
taking 185.624 seconds total on Solaris 10 whereas my next worst system
is taking 84.182 seconds

I understand that there would be some performance hit running on a
virtual server; however these differences seem to be several times
larger than I would have expected.  

I've consulted the admin guide
(http://cran.r-project.org/doc/manuals/R-admin.pdf) I used the provided
make file and the only other configuration change I made was to use
shared libraries based on some of the packages I wanted to use
(--enable-R-shlib)  I did have to use gcc (v3.4.6) as I ran into the
problem referenced in the admin guide about trying to build R with
SunStudio12.2.  This was based on a compiled version of gcc that I had
available to me; so not sure if using a newer version of gcc would
provide significant performance gains.  It seems that most of the
information in the admin guide for Solaris details how to get R to build
from source and not much detail about performance configurations or
tweaks to make it run faster.  I've searched around on the internet as
well as the mailing lists and have not found a lot of information yet
about configuring performance from R on Solaris.

An interesting note is that I was able to build R 2.15.1 from source on
an physical Ultra 27 X86_64 server running Solaris 11 (6GB memory) and
had no problems and performance is comparable to all the other systems I
tested (for example, the standard deviation test only takes about 0.6985
seconds to run on average [gcc v4.5.2])

I'm trying to get hold of a physical Solaris 10 server to evaluate if
Solaris 11 is possibility causing any of the performance gain although I
know that X86_64 will likely perform better than SPARC.

Was wondering if anyone else has experience building R on Solaris or has
thoughts or considerations on what I should look at to increase
performance if possible.

Thanks in advance.


Anthony



        [[alternative HTML version deleted]]

______________________________________________
R-devel@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel

Reply via email to