| Popis: |
NASA’s High-End Computing Capability (HECC) facility in NASA’s Advanced Supercomputing (NAS) division provides considerable resources and services to current and future Science Mission Directorate missions and projects. As part of the support for all the Agency’s Mission Directorates, half of the HECC compute and storage capacity is allocated to the SMD. The FY21 budget for HECC was $47M, which supports a facility with >17,000 compute nodes with approximately 1 million cores, over 100 PB of data storage, and provides for 100 million Standard Billing Units (SBUs) per year. In addition to compute and storage, the funding provides for supplementary analysis systems, scientific consulting for optimization of code and a help desk. HECC also provides data analysis and visualization as a service to users. NASA-funded projects can make use of HECC resources at no direct cost through allocations of SBUs as controlled by the Mission Directorates. In addition to the baseline shares, NASA programs/projects can purchase their own dedicated compute and storage resources at HECC on the margin for as low as $0.09 per SBU and ~$100K per PB, representing a 20X cost advantage over commercial cloud resources such as Amazon Web Services (AWS). While the traditional focus has been on modeling and simulation, the growth in the size and nature of SMD data sets is motivating changes in HECC. Upcoming missions, such as the Earth System Observatory’s Surface, Biology and Geology mission, will be bringing down unprecedented amounts of data (10-15 TB/day) and generating data products at ~75 TB/day, presenting new challenges for data processing and collaboration. NAS is tailoring new file systems to better support large observational data sets, which typically generate many small files and engage in much more random file i/o than the data streams from simulation runs. We are working to improve the availability of HECC through facility changes to support low-latency applications and improve the availability of HECC systems. SMD’s Open Science initiative promises to accelerate the pace of science discovery by making NASA science data sets broadly available to the scientific communities through cloud based DAACs. To meet tomorrow’s data processing challenges we are improving networking and connections to cloud services such as AWS and developing hybrid cloud/HPC computing environments so that NASA can leverage the low cost of HECC compute for future SMD missions. The NAS is supporting several hybrid pilot projects, including HySDS, MAAP, and an urgent computing project for streaming satellite data analysis and event-driven modeling for wildfires. HECC is also supporting SBG’s Space-based Imaging Spectroscopy and Thermal PathfindER (SISTER) pathfinder pipeline study with the development and open sourcing of Ziggy, a science pipeline control infrastructure package that was developed initially for NASA’s Kepler and Transiting Exoplanet Survey Satellite (TESS) missions. Ziggy has been applied to architect a science pipeline to reprocess the 55-TB Hyperion data set collected over 18 years by Earth Observer 1. Using the experience processing the Hyperion data set as a proxy for SBG data, we predict that HECC could process the SBG data to level 2 with reprocessing every two years for on order $6.7M, or about 1% of SBG’s cost cap of $650M. |
