Improving the Execution Time of NWLang
EMSL Project ID
27690
Abstract
NWLang is a software package developed at PNNL, to model the kinetics of cell populations within a spatially explicit environment. To produce these models, NWLang supports the spatially explicit representations of the environment and the organisms residing within that environment. Coupled with the inherent stochastic nature of biological systems, NWLang will be used to simulate systems with computationally intensive model. In addition to the computational needs, the output of these simulations will require a significant amount of storage. These characteristics will require users to run NWLang simulations in a high-performance computing environment.To make NWLang efficient and scalable, the code sequential version of the code has been rewritten to make it more efficient and amenable to high-performance computing techniques. The correctness and performance of these changes must be verified on a high-performance computing system. As the problems targeted by NWLang will contain between 106 - 1012 individual cells, the performance analysis—and therefore the correctness verification—must be carried out on a supercomputer.
Ensuring the correctness of the code could take 500-1000 CPU hours and the performance analysis could take 8000-9000 CPU hours. The storage requirements for these tests will range from 500 GB to 1 TB of offline storage. To accommodate these requirements, the proposal is a request for 10,000 CPU hours on mpp2.
Project Details
Start Date
2007-08-20
End Date
2008-01-11
Status
Closed
Released Data Link
Team
Principal Investigator
Team Members