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

Re: Hardware advice: seeking echoes of running Linux-PC clusters



Jean Orloff <orloff@lapp.in2p3.fr> writes:

> The boss of the lab here asked me to give a presentation about my
> experience with Linux.. We are a high-energy physics lab... Can any
> of you provide me some numbers/weapons for this uneven fight? (I'm a
> physicist, they're the informaticians...)

There are some several web pages about Linux in High Energy Physics (HEP):

               http://hepwww.ph.qmw.ac.uk/HEPpc/.
               http://lars.desy.de/ww/linux-hermes.html
               http://marie.mit.edu/~templon/ (follow linux link).

The authors of these pages are using Linux in HEP so you may want to
contact them.

Your Computing Division are probably most worried about whether
sysadmin will be timeconsuming, rather than what their users might
want :-( . Therefore you'll want to mention the advantages of package
management systems, such as Debian.

Also that Linux distributions include all the standard physics tools,
TeX, emacs, perl, ghostview, etc as *standard*, preconfigured. No
fuss, no muss.

[ CERN is actually developing a general unix package management system
which is largely inspired by Linux package managers. Free software
leads the way! One of the people involved gave a talk at the GNU
Conference in February. http://wwwcn.cern.ch/dci/asis/ . ]

> But the computer division is hostile to linux..

All the high energy physics computing divisions I know are not exactly
embracing Linux or GNU, so you're not alone there!

Their unofficial motto is "Fear change" and they are rarely convinced
by rational arguments.

However there is hope. About 25 years ago a couple of hackers wrote a
new OS to help port their games program and distributed the source
code fairly freely. It didn't come with any support and initially had
a lot of bugs. Despite these problems most high-energy physics
computing divisions now officially support this OS ;-)

So you can take some consolation in that.

My bet is that Linux and GNU will be dominant in physics research in
the future, but it will take time. Probably the best way is to suggest
introducing a small number of machines on a "trial basis" to "gain
experience".

Of course, when people try Linux and GNU software they usually don't
want to go back to proprietary Unices.... but don't tell that to your
computing division!

-- 
Brian Gough

bjg@fnal.gov                          
http://www-theory.fnal.gov/people/bjg/gough.html


Reply to: