Testing new paradigms for biomolecular simulation

Project Information

Discipline

Biophysics (605) 

Orientation

Research 

Biomolecular simulation is a core application on supercomputers, but it is exceptionally difficult to achieve the strong scaling necessary to reach biologically relevant timescales. We are developing a new paradigm for parallel adaptive molecular dynamics and a publicly available implementation. This framework combines performance-leading molecular dynamics parallelized on three levels (SIMD, threads, and message-passing) with kinetic clustering, statistical model building and real-time result monitoring. Even for a small protein such as villin (9,864 atoms), Copernicus exhibits near-linear strong scaling from 1 to 5,376 AMD cores. We are now further testing and optimizing this framework for use on heterogeneous resource platforms such as FutureGrid.

We are developing a software platform that combines traditional HPC and distributed-computing approaches to scale efficiently over many processor cores in heterogeneous architectures. This requires broad interdisciplinary collaboration between application scientists and supercomputing experts. The resulting package will allow any member of the community with sufficient computing power to perform calculations in a transparent and scalable manner.

This award will enable development of a next-generation computational platform to efficiently utilize the increasing number of processor cores for advances in biomolecular simulation. The software will be freely available under an open-source license.

Project Contact

Project Lead

Peter Kasson (kasson) 

Project Manager

Peter Kasson (kasson) 

Project Members

Per Larsson  

Resource Requirements

Hardware Systems

• alamo (Dell optiplex at TACC)
• foxtrot (IBM iDataPlex at UF)
• hotel (IBM iDataPlex at U Chicago)
• india (IBM iDataPlex at IU)
• sierra (IBM iDataPlex at SDSC)
• xray (Cray XM5 at IU)

 

Our project allows effective utilization of both loosely-coupled and tightly-coupled individual resources; we combine individual resources via a message-passing framework for large-scale biomolecular simulation. The advantage of this is that we can make efficient use of a wide variety of hardware platforms. We plan to use FutureGrid as a testbed for running on heterogeneous resources; given sufficient CPU-hours, we can also demonstrate the power of our framework and the utility of FutureGrid on high-impact scientific problems.

We can make effective use of every core we can get. Our framework takes advantage of fast interconnects when we have them but can also work with individual nodes.

Project Timeline

Submitted

04/30/2011 - 19:15