SIE's Christmas shopping
Christmas shopping time has already come, and gone, for the SIE. Following
the directives set by the Head of the Research Division and by the Head of
the Graduate Studies Division, and in close collaboration with the SIC, we
have bought the following hardware:
- 28 Dell's standard PCs (3.0 GHz, 1 GB RAM,
160 GB HD, 16x DVD+RW), to be assigned to new research personnel and to replace
aging Sun workstations.
- 76 17-inch flat monitors, which will be assigned
to PhD students and fellows, and to those postdocs cramped in small offices.
- A new HP laser printer, which will replace the existing lw2, which is developing
some old age ailments.
- Two powerful "burros", each with two Intel's 64-bit
Xeon CPUs, 4 GB RAM, 2 x 250 GB Harddisks, which will replace the old Ultra
10 workstations in the Users Room.
- A brand new Beowulf cluster, see below.
The new Beowulf: entering the 64-bit world
Boy, was it a hard decision! After countless meetings of the five-strong SIE/DSS "Beowulf Expansion Committee" (made less distressing by a thoughtful selection of cookies and candies), where we examined, compared and discussed bids by Sun and by Dell; and after last-minute bargaining with Dell and with Sun sales representatives, finally on Wednesday 9 November at 17.27 we decided to go for the Dell's cluster. Why? Although each Sun's Opteron processor is somewhat faster and more efficient than Intel's Xeon, the "Beowulf Expansion Committee" unanimously
agreed that the configuration proposed by Dell was better, in terms of number
of processors, total RAM memory, and overall disk space, than Sun's. In fact,
for the same money, we got:
- 16 computing nodes (each with two 64-bit CPUs)
+ 1 master node, instead of just 14 64-bit dual-CPU nodes.
- More total RAM memory:
66 GB instead of 56 GB.
- More total disk space: 5.1 TB instead of 4.2 TB.
- More cache memory per CPU: 2MB instead of 1 MB.
- 5-years warranty instead of just 3.
All the relevant information and details about the new Beowulf cluster
will be given in due time.
Kudos to the new organizers of the IAC Seminars, who in our opinion are doing a great job (and, lo and behold, are using the Wiki to plan and schedule seminars, and monitor their attendance). Together, we have thought about setting up a "Computer Seminars" series, with a frequency of once per month or so. The idea is to give short (15-20 minutes) presentations or talks about specific and practical topics, followed by questions and discussion. We have already made up a short list of possible subjects: "Environment variables", "Regular expressions", "Compilation and makefiles made easy", ... However we do encourage all of you to propose new topics and to make comments and suggestions about the organization of these "Computer Seminars", by sending an email to .
With the recruitment of Jorge, our Web officer, the SIE's Web activity has
sped up to a pace never seen before. Among this year's accomplishments we can
- The SIE Forum
- The SIE Wiki site
- The La Palma car booking page
- The IAC preprint submission pages
- The Multimedia Service (SMM)
- The re-designed Research Area Intranet, that we hope to launch soon.
As for Conferences and Research Groups, we built the websites for:
- The Metal Rich Universe
- Ultralow-mass star formation and Evolution
- The Solaire Network
- The IAU Symposium 241 site is still in progress (this Conference is organized
by the IAC and will be held in La Palma in December 2006).
All these sites have been developed following the latest standards and
techniques (as presented in the "XHTML & CSS: standard-compliant web design" course),
and using PHP+MySQL based tools.
OpenSSI experimental cluster in the CCA
While not in the same league as the brand new cluster described above, we are
experimenting with a load balancing cluster for the CCA (codenamed SIAC). For
those of you not familiar with the terminology, there are three main types of
, and High Availability
. The goal of SIAC is to join up a number of machines (initially
four) to balance the load amongst their processors in a transparent way. That
is, users will use the PCs the same way as always, and perhaps they would not
even be aware of using a cluster. The cluster would just give them the impression
of using a machine with multiple processors, which could be used for parallel
applications (e.g. MPI programs) or embarrassingly parallel applications (i.e.
applications made up of a number of independent tasks). To accomplish this goal
we are using the software OpenSSI
. Right now SIAC is in an alpha state (i.e.
not yet fully functional), but if you want to see it working or if you are interested
in learning more about this project, please get in touch with us by sending an
email to .