[Bioclusters] Admins / node?

[Andrew D. Fant]

Brodie, Kent wrote:
>
> Gee- isn't the user support side dropping everything you're doing (say,
> restoring that critical TNS file the DBA accidently deleted), and
> assisting to do whatever the aforementioned omnipotent PI requests?

Actually, I think that an effective cluster group should be well integrated
with either a good help-desk or distributed front-line support organization so
that the user/application support side doesn't have to get tied up with file
restores and can spend more time on technical user consulting 8).

To make more biases clear, my first job was on a help desk for a supercomputer
center in the southwest.  I got to spend a decent amount of time working at
technical user consulting and I was sold.  I also got to see SUNY Buffalo's
Center for Computational Research on site a couple years ago, and I think it's
a model for good user support.  They have a full staff of application experts
working alongside the people who keep the systems running, and they are
encouraged to remain active in their own research fields.  I don't necessarily
think you need a PhD to be useful as a user consultant, but your utility is
limited if you can't think like a researcher who wants to use the system.

And actually, if the PI has the money to run their own cluster (which appears
to be fewer of them now, the NIH and NSF both seem to be increasingly
interested in central research computing facilities in their latest round of
grant proposals), I would suggest the best support model is contracted time
from a central support organization to provide sysadmin services and user
support coming from inside the research group.  I just think that in central
services clusters, the user support needs to come from the organization and not
in borrowed time from the groups using the system.  It's far too easy for a PI
to get jealous of the time their post-doc spends helping other people solve
non-renumerative problems.

> Seriously- I agree with Andy's extended post- very well put.  And yes,
> the system support staff need to know "enough" about the sciences and
> tools to be able to be certain that a simply security patch update
> didn't just hose the heck out of the cluster's ability to create
> accurate results....   That is actually one of the things that makes our
> respective areas quite exciting, educational, and most certainly NOT
> boring like a "standard" IT shop can get sometimes....

Amen, Brother.  I've seen what traditional unix sysadmins do for a living, and
while there are days that I would like to get back to doing science at least
half-time, I have never had a day where I would want to run email for a fortune
500 company or the database servers for <insert local bank here>.   Here,
there's always something new to play with.

Andy (who is really done now)

--
Andrew Fant    | And when the night is cloudy    | This space to let
Molecular Geek | There is still a light          |----------------------
fant at pobox.com | That shines on me               | Disclaimer:  I don't
Boston, MA     | Shine until tomorrow, Let it be | even speak for myself