[R] Snow & rvpm

[vittorio]

Alle 19:03, giovedì 01 dicembre 2005, Martin Morgan ha scritto:
> In the example at
> http://www.stat.uiowa.edu/~luke/R/cluster/cluster.html clusterCall
> divides the parameter R by the number of nodes -- what you've done is
> calculate 999 bootstraps on each node, and compared the execution time
> to 999 bootstraps on one node.
>
> You probably want 'clusterCall' to be smart enough to partition the
> bootstraps between nodes, and then collate the results. It doesn't do
> that.

What program does it?
Vittorio


>
> vittorio <vdemart1 at tin.it> writes:
> > At office, using the internal LAN at my disposal,  I'm having a go at
> > parallel computing - to begin with - with pvm, rpvm & snow.
> > The two boxes are as follows
> >
> > Remote machine uffbsd:
> > CPU: Intel(R) Pentium(R) 4 CPU 2.00GHz (1994.13-MHz 686-class CPU)
> >   Origin = "GenuineIntel"  Id = 0xf24  Stepping = 4
> >  real memory  = 260046848 (248 MB)
> >
> > This machine NbBSD:
> > CPU: Mobile Intel(R) Pentium(R) 4 - M CPU 2.00GHz (1993.54-MHz 686-class
> > CPU) real memory  = 536674304 (511 MB)
> >
> > And starting library snow under R I have the following situation
> >
> > clusterCall(cl, function() Sys.info()[c("sysname",
> > "release","nodename","machine")])
> > [[1]]
> >       sysname       release      nodename       machine
> >     "FreeBSD" "5.4-RELEASE" "uffbsd.myd"        "i386"
> >
> > [[2]]
> >          sysname          release         nodename          machine
> >        "FreeBSD"    "6.0-RELEASE" "NbBSD.myd"           "i386"
> >
> > NOW,
> > using the example of boot in the end of page
> > http://www.stat.uiowa.edu/~luke/R/cluster/cluster.html
> >
> > I find this amazing result:
> >> system.time(cl.nuke.boot<-
> >
> > clusterCall(cl,boot,nuke.data,nuke.fun,R=999,m=1,fit.pred=new.fit,x.pred=
> >new.data)) [1]  0.0078125  0.0078125 27.9609375  0.0000000  0.0000000
> >
> >>  system.time(nuke.boot<-
> >
> > boot(nuke.data,nuke.fun,R=999,m=1,fit.pred=new.fit,x.pred=new.data))
> > [1] 26.976562  0.109375 28.484375  0.000000  0.000000
> >
> > There's not that much gain in time between my cluster and the local
> > computation, isn't it.
> > Now my question is:
> > What could have been gone wrong and what should I verify?
> >
> > Ciao
> > Vittorio
> >
> > ______________________________________________
> > R-help at stat.math.ethz.ch mailing list
> > https://stat.ethz.ch/mailman/listinfo/r-help
> > PLEASE do read the posting guide!
> > http://www.R-project.org/posting-guide.html