[Beowulf] Re: Beowulf Digest, Vol 45, Issue 28

[Mark Hahn]

> We did play with the Pathscale compilers once but the license server
> they used caused us too much of a user support pain as it binds
> licenses to a user for a period after the compiler has been used.

they've basically fixed that by reducing the timeout.  the initial 
timeout appears to have been modeled on an organization with full-time
developers, who would actually keep using the compiler frequently 
for most of a working day.  thus the number of licenses would correspond
sensibly to the number of developer seats.  this makes no sense in an
academic cluster environment, of course, where it's very rare for a user
to spend a long, solid time recompiling.  the current timeouts work very
well for us.  (if there's interest, I can grep some usage stats.)

> If they swap to FlexLM we may revisit that.

I hope not.  flexlm isn't terrible, but it's certainly no better
than Pathscale's license server or other RYO LS's like Allinea.

regards, mark hahn.