A Multiplatform, Cloud-Enabled Mesoscale Model for Business Solutions

In our continuing work focused on providing weather-sensitive business solutions spanning two decades, IBM's Deep Thunder^TM service has provided operational forecasts daily for numerous regions including: New York (upstate and downstate), Vermont, Michigan, Rio de Janeiro and Brunei, among others. The overall configuration has evolved and improved over time to reflect improvements in NWP model capability as well as computational efficiency. The NWP component is derived from a configuration of the WRF-ARW community model. While the deployment of Deep Thunder has traditionally been on a dedicated high-performance computing cluster, recent cloud computing development and production environments have supported other deployment strategies.

We will discuss the development of an updated implementation of Deep Thunder within a new framework. The framework enables numerous additional features and improvements which include an automated build process to support the deployment of the NWP as well as other numerical models on different platforms, including cloud and bare metal cluster environments. Another goal of the work is to understand the performance of high-performance NWP codes in a cloud environment, and in particular the bounds for solving highly localized NWP models in order to support scalable commercial deployments for real-time and on-demand applications.