Traditionally, galactic simulations require super computers (HPC) to run. This is because the effects of forces such as gravity need to be applied to the entire system being modelled; all objects being modelled will exert a force on all of the other objects present in the simulation. The GalDyn group’s research looks at the orbits of particles within galaxies, such as stars and dark matter, and how these are affected by changing a wide variety of orbital parameters. This makes the problem highly parallelisable and ideal for running on the Grid.
Realising the Grid offered the tools required to run tens of thousands of simulation jobs, and manage the large data sets these jobs would produce, the GalDyn group approached GridPP via the Lancaster Tier-2 site. Following the instructions in the GridPP UserGuide, the GalDyn team were setup with grid certificates and access to the NorthGrid Incubator Virtual Organisation to give them access to GridPP computing resources via the Imperial DIRAC instance. Their custom software, written in C++ and stored in a BitBucket repository, was deployed using a CernVM-FS repository hosted on the GridPP RAL Stratum-0, making it instantly available to any UK Grid site. By using a GridPP CernVM, the GalDyn team were able to locally compile and test their software for Grid use.
What they said
“We can run our code on the Grid and are looking to move into production with tens of thousands of jobs required for our final results. It’s nice to know the facility is there and quick to get setup, instead of having to worry how long it will take to get a load run!”A. Clarke, GalDyn
- DIRAC: for Grid job submission and management;
- DIRAC File Catalog (DFC): for data storage and metadata management;
- CernVM: for creating a User Interface to the Grid and software testing;
- CernVM-FS: for software deployment and management.
Supporting GridPP sites