(Posted also to debian-user)
In my office network I have access to a debian powerpc server and 2 freebsd
6 servers (actually one of them is my notebook).
Experienced user of the statistical software R, I have now a go at parallel
computation via (r-)pvm and snow under R to enhance the performance of a
heavy duty statistical problem involving many iteration on the calculation
of models.
I moved my first steps executing the example which comes with snow with
success on the two freebsd boxes. Plain sailing! I can't say the same when
I joined the powerpc debian box to the cluster of PCs.
In a nutshell, I now can add under the pvm shell the 3 PCs and issuing the
mstat hostname command on all of them they're all OK (this either with the
debian box as master and the other two as slaves or with one of the fbsd
boxes as master)BUT it happens that:
Case 1) The Debian powerpc box as master and the 2 freebsd boxes as slaves:
under R I load the snow library but when I issue the command to make the
cluster>cl <- makeCluster(3,type="PVM")
R complains that tid<0 and I cannot go further even thoug R goes on working.
Case 2) One of the 2 fbsd boxes as master and the others as slaves: R and
snow seem to work; I can makeCluster but R freezes as soon as I try to
execute a cluster order like parApply, clusterCall, etc.
What should I check?
Vittorio
