New Jobs section in parallelcrunchers.net

[Robert G. Brown]

> On Sat, Oct 06, 2001 at 12:52:48PM +0000, Pedro Díaz Jiménez wrote:
> 
> > My point is that if we want to have additional services and features
> > for our community we have to discuss what _we_ want and _how_ and
> > _where_ we want it.
> 
> OK, so I'll start:
> 
> I'd like to see "discussions". These "discussions" ought to have a
> nice, searchable archive, and they should fit nicely with the existing
> discussions, mainly this mailing list. [ Pedro said the same thing in
> his email. ]
> 
> I'd like to see a jobs site. But I think it ought to be linked with a
> general job site for more traffic, i.e. "use this job site, be sure to
> mention 'linux cluster' in your resume or job description." Preferably
> the jobs site ought to be both free to put a resume and put a job
> in. Most existing jobs sites are free to list a resume, but charge
> employers a fee.
> 
> I'd like to see a listing of clusters. So far topclusters may be the
> best; they have real humans to review the entries, etc. So is
> topclusters missing any features? I can't think of any.
> 
> I'd like to see a hierarchical, categorical list of links: software
> systems, hardware info sites, whatever. Maybe using ODP (dmoz.org)
> would be good, since they have a submission/editor system that works
> well, they automatically mark dead links, etc.
> 
> It would be cool to have a categorized list of "testimonials", where
> people talk about what clusters they built and why. Then if you want
> to prove to your boss that a cluster might be a good idea, you're more
> likely to find a relevant example. Mandrake has a site for mandrake
> testimonials, but I think we should have one for clusters.
> 
> Anyone else have a comment?
> 
> greg

Pedro,

All of these suggestions are great, but the usual problem isn't having
good ideas or needing discussion to generate more, it is having someone
with the time and energy to make it all happen.  As somebody (Greg?)
already noted, a lot of cluster resources are already available via the
main beowulf link, the beowulf underground link, and a few other beowulf
and cluster and HPC sites.  Scyld has humans being paid money (I hope)
to take care of the beowulf site; Clemson has faculty and students
involved who do theirs.  To build and maintain a really good beowulf
resource site is a virtuous undertaking, but it takes a lot of TIME to
build it and debug it, and then still more time to maintain it and groom
it.

Website building is a sort of Darwinian process.  If you have the time
and energy to build a really good beowulf/cluster resource site, you
should just dig right in and do it.  Greg's list is all great, and it is
easy to think up more useful stuff -- a beowulf-specific sourceforge
site (to make it relatively easy to nucleate development efforts around
tools), a large master cluster database with the ability to generate a
variety of online reports (which should have fields for the testimonials
Greg mentions, directly connected to a database description of your
cluster -- number of nodes, hardware architecture details, application
details, and more).  By putting all of this in e.g.  mysql, one can do
complex searches -- finding a list of annotated clusters running e.g.
dual Athlons AND (gig ethernet OR myrinet) would be fairly simple.

If you build this and take care of it, you won't have any trouble
convincing people to visit it and use the services you provide.  If you
don't REALLY have the time and energy and creative vision required,
well, time will prove that as well, one way or the other.

This isn't intended to disrespect your effort; on the contrary, I wish
you the best of luck.  The beowulf underground site started out a few
years ago as little more than an announcement, and today it is one of
the most useful cluster sites around.  I'd be happy to see more, useful
sites.

   rgb

-- 
Robert G. Brown	                       http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567  Fax: 919-660-2525     email:rgb at phy.duke.edu