Project
IBM Research Homepage 
 Research Home  >> Computational Science Collaborations with IBM Customers



Collaborations with IBM SP Customers - University, Industry and Government

The group has active ongoing collaborations with IBM SP customers at Universities around the world, IBM Industrial partners and Government organizations to develop new research ideas in Biology, Weather Modeling, Algorithm Design, Software Development etc. The aim of this effort is not only to do high quality research but also to create an awareness of IBM's software and hardware by developing scientific collaborations with the goal of helping IBM customers develop efficient code and algorithms on the IBM SP.

Johns Hopkins University
At Johns Hopkins University, we are collaborating with the Center for Imaging Science to determine a metric to judge the deterioration of the hippocampus, an organ in the brain associated with diseases such as Alzheimer's. The method involves using Computational Fluid Dynamics ideas to find the metric distance between 'normal' and 'patient' hippocampii and then using this metric for possible early diagonosis of degenerative diseases. This research is being done in the group led by Professor Michael Miller on the new IBM SP supercomputer at the recently established Whittaker Biomedical Engineering Institute at Johns Hopkins University.

Also at Johns Hopkins, we are working with Professor Raimond Winslow, Director of the Computational Medicine & Biology Laboratory on modeling and imaging electrical activity of the heart using the IBM SP. The algorithm used for the modeling was implemented on the IBM SP by researchers at TJ Watson lab working in collaboration with students and faculty at the Winslow lab. An outgrowth of this work was a project to model Ca gain and gradedness in heart muscle cells, which has led to interest in setting up a joint collaboration between Johns Hopkins University and the IBM Computational Biology Center to do whole cell simulations.

Geo Fluids Dynamics Laboratory
We have helped implement a parallel version of the GFDL Hurricane Modeling Code on the IBM SP at the National Weather Service/NOAA. The code is developed and maintained by GFDL and uses triple nested, moving grids to predict the path of hurricanes in the Gulf of Mexico. Originally optimized on CRAY computers, the code now runs highly efficiently on the IBM SP and is capable of modeling a 72 hour forecast in less than 35 minutes on 42 WH-1 processors. It is slated for operational use by the National Weather Service/NOAA for hurricane forecasts in the Gulf and along the eastern seaboard during the next hurricane season and beyond. We are now helping to couple this code to an ocean modeling code for a more reliable modeling of the physics of energy transfer between the atomsphere and the ocean, which will lead to more reliable forecasts.

 Privacy | Legal | Contact | IBM Home | Research Home | Project List | Research Sites | Page Contact