The continued rise in computational power is significantly enhancing
medical research and promises to expedite the process of translating
research in basic science to the clinic. High throughput
computational screening as well as highly intensive biomolecular
simulations are accelerating the process of drug discovery to a new
level. Blessed with the nation's second largest School of Medicine,
as well as a rich high performance computing cyberinfrastructure,
such activity is a natural fit for IU.
Showcasing this is work by Dr. Samy Meroueh and his group in the
IU School of Medicine's Department of Biochemistry and Molecular
Biology, along with the Center for Computational Biology and
Bioinformatics. He is using IU's Big Red supercomputer in research
aimed at the discovery and design of small molecules that can be
used for the treatment of various diseases. Big Red allows Dr.
Meroueh to multiply many times the rate at which he can identify
potentially useful molecules, a process that involves screening
millions of candidate compounds and computing how well the
candidate molecule shuts down the disease causing agent. His work
has led to the discovery of highly potent agents that block
processes that tumors must undergo to metastasize. Dr. Meroueh is
now working in collaboration with various faculty at the world-
renowned Simon Cancer Center at the Indiana University School of
Medicine to assess the effectiveness of these molecules in animal
models.
For more information, please visit
http://compbio.iupui.edu/meroueh
Indiana University Cyberinfrastructure News
-
-
Organizing committees for the 2008 IEEE eScience Conference are now
accepting papers and proposals for tutorials; posters, exhibits, and
demos; workshops and special sessions on topics related to eScience,
grid, and cloud computing. The conference is being hosted by Indiana
University in partnership with Microsoft Research and will take
place on December 7-12, 2008 at the University Place Conference
Center in Indianapolis, Indiana.
For more information on topics of interest, submission guidelines
and deadlines please visit the conference Web site at
http://escience2008.iu.edu/ -
A recent update to Big Red's LoadLeveler resource manager includes a
new job type: MPICH. LoadLeveler jobs of type MPICH, as opposed to
"parallel", automatically generate host files for use with mpirun's
"-machinefile" switch.
A job's host file can be referred to via the new $LOADL_HOSTFILE
variable; use of the llmachinelist command and subsequent cleanup
are no longer necessary!
In addition, the variable $LOADL_TOTAL_TASKS may be used with
mpirun's "-np" switch. $LOADL_TOTAL_TASKS is calculated from the
LoadLeveler job script keywords "node" and "tasks_per_node". As a
result, changes to job geometry only need to be made with these
keywords; the mpirun command's arguments will be modified
automatically.
Finally, the MPICH job type provides support for some additional
MPI application enhancements which we intend to implement in the
coming weeks. We encourage all of our MPICH users to modify their
job scripts accordingly. -
Beginning June 1, 2008, files that haven't been accessed in more
than 180 days will be automatically removed from IU's General
Parallel File System (GPFS) scratch file system.
The GPFS scratch file system is accessible from Big Red, Quarry,
and Libra.
Please be sure to back up important data stored on GPFS since it is
a temporary scratch file system and is not intended for long-term
storage.
For information on backing up data to the Massive Data Storage
Service (MDSS), see the KB article "At IU, how do I use SFTP or SCP
to access my MDSS account?" at
http://kb.iu.edu/data/avax.html -
When you submit a job to one of IU's supercomputers, there will be
hundreds of jobs in front of yours and more lining up behind.
Optimization ensures all these jobs will run as quickly as possible.
How can you simply and quickly optimize your program? Let the
compiler help. By using three types of compiler switches - machine,
usage, and optimizing - you can make your program fast and sleek
without a lot of work. In this article, we'll look at optimizing
switches only, with a C program compiled using Big Red's IBM
compilers as an example.
Not all switches improve a program's performance. Some are neutral,
some will help or hurt based on combinations, and some will slow a
code down dramatically on their own. Safe bets include -Q to
replace function calls with the function's source code, and -qunroll
to unroll loops as if they were sequentially written source code.
Those that should be used with care include -qhot for combining
loops.
The -O switch is a set of macros for these types of smaller
optimizations. This switch optionally includes a digit (e.g., -O3)
which increases optimization risks as it increases. To offset this,
IBM offers the -qstrict switch, which tells the compiler not to
take risks if there is some doubt. A good first test of our sample
C program might be:
xlc -q64 -O3 -Q -qunroll -qstrict -qarch=ppc970 -qtune=ppc970 \
-qenablevmx -qaltivec -o my_project my_project.c -lm
Every program is unique, but a good compiler can typically improve
the performance of even a well-written program by 1/3. Reading your
compiler's man page and experimenting with a few simple optimization
switches may even cut your job's run time in half.
IU's High Performance Applications group may be able to help with
compiling and optimizing your programs. For more information, see:
http://rc.uits.iu.edu/hpa/index.shtml -
Monday, May 12, 11:00am-1:00pm - ICTC Main Lobby, IUPUI
Learn about IU's supercomputers from the experts!
You're invited to stop by and visit with the system administrators
who run IU's supercomputers, Big Red and Quarry -- two of the most
powerful supercomputers in the world.
You're welcome to ask questions, share ideas, or just chat in an
informal setting. This is an opportunity for you to learn how
these centralized research computing resources can help support
your work.
--------
Thursday, May 29, 12:30-1:30pm -
IUB: Radio-TV Building Room 180
IUPUI: ICTC Room 497
Live URL: mms://wms.indiana.edu/rt_round_table
Research Technologies Round Table
Science Gateways
Marlon Pierce
After more than a decade of development, tools for Web-based access
to computing resources and data archives are now very mature. This
month's Roundtable will include discussion of these Science Gateways
and the Grid middleware that they access, architecture and standards
used by the science portal community, component-based Web portals,
Web Services, and workflow (or service orchestration) tools. Also
discussed will be Web 2.0 and Cloud Computing approaches to
resource and data access, and these tools' eventual merger into
Science Gateways and portals.
Please note the new Radio-TV building location for IUB.
Videoconferencing equipment will broadcast to the IUPUI location. -
The maintenance window for all systems is the first Tuesday of each
month, 7am - 7pm EDT.
Outage reports are available online at:
http://racinfo.indiana.edu/hps/research/bigred/outages.shtml
http://racinfo.indiana.edu/hps/research/libra/outages.shtml
http://racinfo.indiana.edu/hps/research/quarry/outages.shtml -
If you have questions pertaining to IU's cyberinfrastructure, or you
are encountering some difficulty, there are several ways to obtain
help.
The IU Knowledge Base (http://kb.iu.edu) is an excellent source of
help on how to do things.
An introduction and overview titled "Indiana University's
CyberInfrastructure: The least you need to know" is available at
http://rtinfo.uits.indiana.edu/documentation/
For more information, go to:
http://rtinfo.uits.indiana.edu/
http://racinfo.indiana.edu/hps/
If you have problems which the KB does not enable you to solve,
questions about system outages, or if you just have a problem and
you don't know who to contact, send email to
researchtechnologies@iu.edu.
