Hardware we use to run the models
Some of the operational models are run on our dedicated 552-core linux cluster. Others utilize 1024 cores on Google cloud every day. We are gradually migrating more of the daily runs to the cloud.
The raw NWP output requires extensive further processing to produce weather maps, meteograms, archives, verification statistics, and post-processing forecast improvement. To accomplish this, we run an additional 28 unix servers (= 324 cores) for file/web/ftp serving, RAID storage, tape storage, and database archives. For the students, postdocs, scientists, and support staff, we provide 25 high-end workstations (=100 cores). For field work we have 5 Durabook laptops.
In addition to the daily real-time runs that must be completed in time to meet BC Hydro deadlines, we also make a significant number of experimental/research/exploratory runs for which timeliness is not an issue. For this research, we use some of our slower servers, unused interstices in our large-server schedules, and we make extensive use of Compute Canada resources. Over the past several years, we have won the following Compute Canada Research Allocation Competition (RAC) awards:
- 2018 - Awarded 92 core years & 50 TB storage
- 2019 - Awarded 90 core years & 50 TB storage
- 2020 - Awarded 107 core years & 50 TB storage
- 2021 - Awarded 80 core years & 50 TB storage
- 2022 - Awarded 80 core years & 50 TB storage
We also use a lot of Compute Canada/WestGrid computing via the normal user queue.