Oppositely, the research and academic community code targeting the highly heterogeneous computations that needs to be portable, interoperable and flexible enough to scale from the developmental laptops to the pilot model sizes, then to sustain the testbed, operational and to then ultimate stress loads.
Several years ago, during organizational transitions between multiple HPC platforms, NCEP's Environmental Modeling Centers (EMC) initiated the development and deployment of the basic principles of the flexible Functionally Equivalent NCEP operational environment (FEE). The solution was suggested with a 1) deployment of the conventions in merging versions on the NCEP developed libraries, 2) full parametrization of the HPC environmental variables in operational models, scripts and building systems 3) unified vertical structuring of the NCEP models, libraries, operationally utilized third party libraries and services within EMC NCEPLIBS group 4) designing model-driven hierarchic HPC modulefiles, to able to equip code version control with the computational FEE footprints, and 5) introduction of FEE verification system, containing both verification of the models accuracy and FEE availability.
Other words, the development of EMC in-house operational standards, FEE conventions, parametrized NCEP control and builds verification methodology, as well as FEE the naming and structural standards should bridge R2O collaborators to adapt or interface with NCEP formal requirements and use mostly the same code and building systems, migrate between versions of the depended libraries, and also to verify an availability of such NCEP FEE components on very initial phase of the porting.
This talk is targeting to share a progress of the OPS and R&D deployed FEE components, their interactions and EMC experience with OPS accepted standards of upcoming NCEP operational implementations.