|Perl: the Markov chain saw|
Large Windows network: One 32 bit perl or 32 and 64 bit perls?by Boldra (Deacon)
|on Jan 08, 2010 at 10:09 UTC||Need Help??|
Boldra has asked for the
wisdom of the Perl Monks concerning the following question:
We're looking at modernising our standard perl install on 1000+ Windows machines of mixed versions (2000 through 2008, 32 and 64 bit, various hardware, including VM).
We are considering two options:
1. One 32-bit perlWe build a 32 bit perl and modules, place it on a filer (eg \\ourperl\bin\perl.exe), and always call that one perl for everything. When we need a new module, we test it on one 32 bit and one 64 bit machine locally and then update the master if the tests pass.
CON: Some modules running on 32-bit perl are buggy on a 64-bit os (eg Win32::TieRegistry and Win32::OLE)
2. 32 and 64 bit perlsWe keep two perls up to date, eg \\ourperl\32bit\bin\perl.exe and \\ourperl\64bit\bin\perl.exe, each with it's own versions of any cpan modules we need. Updating modules works similar to 1, but we do the tests and rollouts seperately.
CON: Each machine needs to keep track of the location of perl.
Have I missed anything? Is there a significant performance difference with 64 bit? Is it worth the admin hassle of having all the perls locally installed? We are tending towards strawberry (because of better cpan support, and I'm fairly sure we can compile in INC to point to a network drive).
We also have to consider that on some of these machines other perl versions may be installed, so we have to be careful about changing anything in the global environment (that's what the IPCC said).
Our usage is quite varied, from gui apps called from network shares on user request, scripts started remotely via ssh, and locally started scheduled tasks.