Are we alone in the universe? Could life exist on other planets? Questions like these have inspired countless science fiction stories, but at the University of Exeter's Astrophysics department, these queries are at the forefront of research. The department is using a range of supercomputers coupled with its own bespoke open source software to study how stars and planets are formed.
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
Running computer simulations of how planets and stars form in different parts of the universe and under many different conditions should give scientists some indication of what types of planets and stars they could find in the further reaches of space.
"We want to know where the human race comes from and how many other solar systems are out there," said Matthew Bate, professor of theoretical astrophysics. "Once we understand a little more about star formation, we can hopefully go forward and make predictions about how the process might vary in different parts of the galaxy."
The department has written its own computer program to simulate how stars are formed. It then uses a powerful computer to run a detailed simulation of a star formation. However, the department needs to run simulations of star clusters (several stars) to get an idea of how accurate the model is. This has increased the computing power the department requires.
The project uses 1,280 computer cores, which is equivalent to the processing power of about 300 quad core desktops linked together. The department is networked using an infiniband communications link.
Infiniband allows very quick passing of information packets between processor nodes, up to 96Gbps. It includes quality of service and failover, and is a key requirement for the project.
"If you're trying to model a cloud simulation, then each part of that's cloud formation could be carried out by a different processor. Because one part of the cloud's formation is dependent on another part there is a need to pass information between each different process very quickly, which is why we use this network fabric," said Bate.
The department has also been forced to install its own supercomputer to carry on research following the closure of the computing resources at the UK Astrophysical Fluids Facility. Funding for the facility ended in 2007 and it was is no longer able to provide a computational resource to the UK astrophysics community. A statement on the facility's website said it appears unlikely that UKAFF will receive further funding.
In January, the university took delivery of a 1152-core SGI Altix ICE 8200, based on Quad-Core Intel Xeon 5400 Series Processors
"With the new computer in place we can run between 100 to 1,000 simulations at once which gives us very good statistics. It also allows us to run more precise simulations."
The entire project runs on SuSE Linux Enterprise Server (SLES) - a Linux distribution supplied by Novell. It is targeted for servers, but can be installed on desktop computers for testing as well. SLES receives much more intense testing than other SuSE Linux products with the intention that only mature, stable versions of the included components will make it through to the released product.
"Astronomers tend to use Unix or Linux-based operating systems for their work, pretty much exclusively. A lot of our code tends to be written for those operating systems and we don't write code for Microsoft Windows," said Bate.
Surprisingly though, the actual code that Bate has written to perform the analysis on the simulations is written in Fortran, which could be run under either Windows or Linux. But because his program has to interact with other analysis packages that are Linux based, this has dictated its use as the de facto operating system.
Perhaps the most interesting part of the project is the fact that it uses no database application even though it generates terabytes of data. Last year alone the project generated 3-4 terabytes of data. The new supercomputer the department uses has 30 terabytes of disk space.
"You're talking about very large data sets. We tend not to store them for very long. We'll run the simulation and keep the data for a year or so on DVDs," said Bate.
The other key aspect of upgrading its IT infrastructure was the need to keep the computer cool. There wasn't any space left to install more air conditioning, so it was imperative to build a cooling module inside the computer. The next thing the university is looking to install is something called OpenMP.
OpenMP (Open Multi-Processing) is an application programming interface (API) that supports multi-platform shared memory multiprocessing programming in C/C++ and Fortran on many architectures, including Unix and Microsoft Windows platforms. An application built with the hybrid model of parallel programming can run on a computer cluster using both OpenMP and Message Passing Interface (MPI).
A key milestone ahead for the project this year will be when it runs a simulation of the collapse of gas containing 500,000 times the mass of the sun. As the temperatures in space increase, it looks as though computing power is keeping pace, too.