Javier Ganivet's work at the SIE
During the last 4 months Javier Ganivet has enjoyed a collaboration grant
with the SIE, which ended on the 30th of November. His work (which has been
summarised in a
PDFdocument)
has been centered around our
Chimera cluster,
focusing in the following areas: recopilation of information about the EM64T
extensions (Intel's effort at 64 bits computing); running a number of different
benchmarks and testing the Chimera cluster with different configurations of parallel
libraries; and configuring the Ganglia monitoring software in order to get more accurate
information about the real CPU load of the cluster. We also remind you that if you have
any particular opinions on how to make the best use of the cluster, you can make your point at the
SIE Forum.
Code running 117 times faster thanks to our Beowulf cluster!
The SIE installs, maintains and administers a part of the IAC's Supercomputing infrastructure (the
Chimera cluster,
and the
Condor grid), but we also
give support and help you make good use of it. As a recent example, we are collaborating
with the "Alta Resolución en Física Solar" group to run the
MOMFBD code (a Multi-Frame Blind Deconvolution
algorithm developed at the Institute for Theoretical Astrophysics at the University
of Oslo and the Institute for Solar Physics at the Swedish Royal Academy of Sciences)
in the IAC Beowulf clusters. As a result, the running time of the code with a simple
set of images went down
from 39 hours using one workstation to around 20
minutes! If you have (or would like to have) code that takes ages to run,
get in touch with us and we might be able to help you save some precious time.
Treat yourself to an "Elastic Cluster" at
Amazon this Christmas
Tired of boring Christmas presents? Amazon has a lot to offer, including an
"
Elastic Compute Cloud"
which you can turn into a Cluster that can grow or sink on-the-fly as needed.
But why would you want one of this? Imagine the following scenario in which
you have an instrument that every night produces some data that needs to be
processed by the following night. The amount (or the complexity) of the data
can be different every night, so you don't know in advance how much CPU power
you will need to process your data. If you have a similar scenario the
"Elastic Cluster" could be a solution for you: every day you can
determine how many CPUs you will need to process the data and dinamically
create your cluster. If this sounds interesting, do get in touch with us to
see the demonstration of the "Elastic Cluster" that we are working
on.
tree and fitscut
We have recently installed two small but useful tools in Linux.
tree is a recursive directory listing program that produces a
depth indented listing of files. Just type tree ~/ to list your home directory.
Type tree --help for the full list of options.
fitcuts is designed to extract cutouts from FITS image format files,
it supports FITS, PNG, and JPEG output types. For usage instructions and relevant
links, check out
SINFIN's Fitscut page.
linklint, program to check Web links
Are you currently developing or maintaining a website? Have you ever looked
for and fixed broken internal and external links? Regularly checking links on
a website should be a staple activity of any Webmaster. Internal links may
break because Web pages are moved or renamed; external links may get broken
because websites are relocated to a new place or server, or the URL address
has changed, or sites have shut down altogether. We have installed
linklint,
a relatively easy to use tool to check both internal and external links. Its
basic usage is described in our most recent
SIEpedia
article.