[Beowulf] OS for 64 bit AMD

Mark Hahn hahn at physics.mcmaster.ca
Mon Apr 4 08:51:09 PDT 2005


> course these are not the only changes.  The big issue was the stack size 
> change.  That one killed off my wireless driver and wreaked havoc with 
> my graphics on my test machine.

but with an srpm and 2 minutes, you could have turned off this feature.
I guess it boils down to which features you are willing to tweak.
some people think that changing system file layout is unacceptable
in a "production" release (they mean "bug-fix-only").

> >    What is much more important in a true "production"
> >    environment is the length of time one can expect to
> >    obtain patches for the OS. No "production shop" that
> 
> I get the feeling that this is an unwinnable argument.  One person 
> (Mark) argues that support patches are effectively a tool to lock you in 
> (and if I characterized this wrong Mark, please feel free to correct 

I argue that the concept of "supported" or "qualified" systems
belies an assumption that components do not conform to a standard (ABI)
or that the standard is inadequate.  customers do not typically 
realize that when someone says "we support this package only on 
RHEL 3.1, ia32, xfs, nvidia", they are *not* stating that their 
package is believed to conform to LSB and use standard OpenGL.
the customer must now either use the exact config, or take the risk
that the package does depend on undefined behavior.

whether any of this matters to one's definition of "production"
depends on your own risk aversion and level of skill in working round
the package's aberrant behavior.  such as 4k stacks.




More information about the Beowulf mailing list