Teleconference Minutes 4 Jan. 2001

Minutes of a teleconference & meeting regarding Condor software 

4 Jan. 2001.

TEAM MEMBERS PRESENT: Churchwell, Kobulnicky (acting secretary), Whitney
TEAM MEMBERS BY TELECON: Wolff, Wolfire
GUESTS: Marilyn Meade and Brian Babler (UW data analysts)
        Miron Livny (Condor PI) with Condor staff Ed ??, Pete ??

Condor web page: www.cs.wisc.edu/condor

Miron described the setup and implementation of condor for the group and
fielded questions.

Disk space: we estimated we'd need ~100 GB of disk storage which is no
issue to them to handle on a TEMPORARY basis.   We'd still need to
purchase our own for permanent storage.  If their resources can't
handle the data volume, then there's the option of purchasing disk
space to put on other machines.

Condor is a daemon that runs presently on unix platforms which allows
nearly transparent distributed computing of large jobs, potentially to
nodes all over the world. They are working on porting it to Windows
NT.  It would work by automatically moving image files and software
executable files from a home location to an under-utilized workstation
with the resources to perform out work.

We would not be able to use IDL to process our images without having a
license for each machine.

The condor Daemon runs as root, while the computational jobs themselves
run either as an existing user on the machine, or as user "nobody"
through set-UID

The security issues of having a root-run daemon have yet to be
investigated, though they expressed no concern about security.  [Chip
wonders if they have had a team try to exploit holes in the condor
daemon...we might ask about this given Stephan's security-conscious
nature.]

Condor has some difficulty crossing firewalls, but Mike thought that
our recently-installed firewall could be configured to allow the
necessary access.

If the executable software does not reside on the compute 
node already, then we must transfer the executable/source code
along with our data to the node which will do the computing.
Thus, we may have to port our data processing code to different
operating systems/architectures.

It's possible that just having Condor on astronomy machines might be
enough to serve our purposes since there are so many workstations around
here that go under-used....but would the workstation owner go for it?

University of Victoria astronomy group has Condor running, we could
call them and ask about their experience.

It is remotely possible that the Condor group could lose funding in
which case they would give us the source code so that our program could
continue.

ACTION ITEMS:  
Some discussion with Stephan will be necessary to
determine possible hitches in Condor installation in the department.

to Minutes index page