Tim Pugh is the supercomputer program director at the Bureau of Meteorology Australia
In this video from the DDN User Group at SC17, Tim Pugh from the Australian Bureau of Meteorology presents: Our Move to a new Data Production Service.
The Bureau of Meteorology, Australia’s national weather, climate and water agency, relies on DDN’s GRIDScaler Enterprise NAS storage appliance to handle its massive volumes of research data to deliver reliable forecasts, warnings, monitoring and advice spanning the Australian region and Antarctic territory.
The Bureau intends to use DDN’s GS14KX to support its new data-intensive computing applications with integrated workflows to the Cray XC40 HPC environment for weather forecasting. We will also consolidate workflows from multiple legacy systems into a high-performance, replicated storage system,” said Tim Pugh, supercomputer program director at the Bureau of Meteorology Australia.
With DDN’s leadership in parallel file systems at scale and its deep expertise in Lustre and IBM Spectrum Scale environments, DDN is well positioned to support weather and climate organizations as their unabated data growth continues and as they require acceleration technologies such as flash native caching to further speed simulations and hot data computations. For example, DDN’s Infinite Memory Engine solution can accelerate performance speeds by 3x and make application completion times predictable.
Technologies such as DDN’s flash-native storage cache – Infinite Memory Engine – are boosting weather code performance to process more data, faster. For example, researchers at Ireland’s high-performance computing center, ICHEC, realized a 3x performance boost of the popular Weather Research and Forecasting (WRF) model, with no code changes and with one-tenth the required infrastructure when using Infinite Memory Engine. With this type of accelerated performance, supercomputers can provide a quicker turn time for atmospheric and ocean simulations so that severe weather events can be predicted with sufficient time for preparedness. More performance also allows for better fidelity, with grid sizes reduced to 1 to 2 km on the more granular models. Improved fidelity translates to more accurate forecasts, so localized phenomenon such as tornadoes, hailstorms, and intense downpours can be predicted at more useful scales.
