Project reference: 1518
Students will become acquainted with user friendly HPC tools: profiler, grid middleware (ARC), visualization tools (VMD, VirtualGL/TurboVNC) as well as one application from the PRACE community code repository: Siesta.
About the tools
We employ ARC middle ware and SLURM for HPC jobs on all of our HPC sites. The ARC middleware provides an opportunity for the user to submit jobs remotely from a local machine. This makes the usage of HPC machines more user friendly because the user do not need log in to the HPC frontend machine.
We have three visualization machines. The GPU card of the visualization machine can be accessed with VirtualGL/TurboVNC pair. With VirtualGL the OpenGL commands are redirected to the graphics accelerator and the rendered images are sent back to the client machine. Students will visualize input and output data of the selected material science applications using the VMD visualization package.
About the application
Siesta is density-functional based packages for electronic-structure calculations Both codes enables the calculation of nano scale structures at the atomic level. The packages are written in Fortran and utilize MPI for parallel communication. Students have to compile and optimize each code as well as write job scripts for the scheduler. They will run some selected examples, from small molecules to large bulk systems to measure and check scalability and determine the optimal resource requirement. They will visualize input and output data with VMD as well as plot scalability graphs for performance analysis.
Project mentor: Gabor Roczei and Tamas Hornos
Site Co-ordinator: Tamas Maray
The student will learn how to compile optimized Fortran codes with MPI (OpenMPI, SGI-MPT), profile and measure scalability of an application, run Siesta on HPC systems, manage HPC jobs with ARC middleware and SLURM, usage of visualization tools (VMD, VirtualGL/TurboVNC)
Student Prerequisites (compulsory)
Basic UNIX user knowledge.
Student Prerequisites (desirable)
UNIX shell scripting and basic knowledge of quantum physics.
- Week 1 : Training
- Week 2 : Introduction to Siesta, VMD, VirtualGL/TurboVNC, Open Grid Scheduler, ARC middle ware
- Week 3 . Plan (Student will define specific tasks in steps to completion)
- Week 4 – 7: Explore visualization methods
- Week 8: Finalize report
Final Product Description
Two visualization outputs will be:
- Scalability graph (comparison the CPU numbers versus the number of atoms)
- Visualization of the input and output geometries and volumetric data like (eg. electron density)
Adapting the Project – Increasing the Difficulty
Adapting the Project – Decreasing the Difficulty
- HP clusters
- SGI ICE cluster
- Nvidia Quadro FX5800 based visualization subsystems
- Nvidia Tesla M2070 GPU cluster
National Information Infrastructure Development Institute