[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Condor-users] Pool Of Virtual Boxes

Hi Kevin:

On Fri, 2011-04-01 at 11:09 +1300, Kevin.Buckley@xxxxxxxxxxxxx wrote:

> The apparent lack of (mis)interaction with the host system is
> interesting.

I am sure that the Vbox is affecting our machines in some way (the
memory usage probably being at the forefront). However, povb is
installed on undergraduate lab machines where students don't know that
we are running a VM in the background. It seems that naive users who
aren't looking for a performance impact don't feel like the system is
running slower than normal. 

> Would PoVB (I'm sure it should be a small "o" in "of") allow for
> a one-off installtion of a VM by our central facilitator over
> which we, in the Comp Sci School, would then have the ability
> so as to facilitate the needs of the grid users.

This was the reason for designing povb. Create an installer that Windows
sysadmins could use and then let the researchers decide how to use the
facility (via condor configuration and Linux OS config). The images that
are provided on sourceforge are generic - they are meant to be
customized to suite the needs of the researchers (by installing
additional software, etc).

> Furthermore, given that some users have become used to recompiling
> the UNIX codes with Cygwin so as make use of the existing facility,
> what happens when/if you have a Condor Grid of PoVB VMs using the
> same hardware as a grid running on the Windows hosr?

While we have not looked specifically at this use-case, I don't see any
problem with running the two systems side-by-side (noting that the
Windows-Cygwin system would always "win" over the vbox system). Since
povb is, by default, monitoring non-povb cpu useage, the povb slots
would go into the owner state when a cygwin job is running. The only
downside, as noted by others, is that you need to assign some amount of
memory to the Vbox machine, which could impact "large memory"
cygwin/windows jobs.