ESRL Global Systems Division
NOAA's Supercomputer at FSL Becomes More Powerful
With winter forecast experiments in full swing, the Forecast Systems Laboratory welcomes the recent upgrades to the NOAA High Performance Computer System at the FSL campus. This new system, built by High Performance Technologies, Inc. (HPTi), was designed using the same principles as the previous systems, which maximize the price and performance one can achieve today using cluster technology. The system consists of 630 individual 64-bit Intel processors, "operating like a symphony to provide better forecasts," says FSL director Dr. Sandy MacDonald. The combination of these high-performing subcomponents significantly raises the volume of work that can be completed in the same time frame as before. Though this system has only half as many processors, it far outperforms its predecessor, with the ability to access, process, and distribute much larger datasets. The 50-terabyte (TB) storage system is three times larger than the replaced composition and twice as fast.
This timely upgrade offers NOAA more compute power to accommodate many new projects that require the fastest computers available to support its mission of improving weather forecasts through higher resolution prediction models. For example, the Rapid Update Cycle (RUC) model is being tested at a resolution that is almost twice that currently run in operations. Also, FSL and its partners are involved in the Developmental Testbed Center's Winter Forecast Experiment currently running a real-time version of the Weather Research and Forecasting Nonhydrostatic Mesoscale Model (WRF-NMM) that requires almost 500 processors, or one-third the total of the previous system. These models are running two to three times faster on the new system, and generate forecasts in less than half the time before the upgrade, freeing up researchers and forecasters for other pressing tasks.
Installation of this advanced computer technology allows FSL and its customers, especially the National Weather Service, to better evaluate and demonstrate improved weather prediction. For more information about this system, see the FSL High-Performance Computing System Web site.
Name: Leslie B Hart