Tuesday, 14 January 2020: 10:30 AM
155 (Boston Convention and Exhibition Center)
We present plans and status of our efforts to gather and analyze requirements for computing, data and software capabilities needed to carry U.S. Navy operational weather prediction through the next decade. The analyses will inform decisions on capability, capacity and architecture in HPC acquisitions, both installed and cloud-based. Factors affecting performance, scaling and efficiency include computational intensity, power efficiency, memory bandwidth and latency, offload device overheads, network performance and I/O. Analyses are based on projected workloads comprising high-resolution deterministic and ensemble configurations at forecast scales, coupled earth-system simulations at seasonal to annual scales, cycled data assimilation, and projected analysis workloads including AI and machine learning.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner