1. On Advanced Scientific Understanding, Model Componentisation and Coupling in GENIE Sofia Panagiotidi, Eleftheria Katsiri and John Darlington

2. 1 Motivation Reverse engineer GENIE to inform research on component coupling frameworks:  Analysis of a monolithic piece of code such as GENIE.  Extraction of interface and model coupling semantics.  Development of methods for capturing such semantics and validating the development.  Methods to express at higher-levels co- ordination structure.

3. Goals  A flexible coupling framework where components can be individually developed and coupled together.  The coupling needs to be correct, intelligent and without undue programming effort.  In synch with existing efforts: (GCF, Iceni I, Apples AMWAT, Reality Grid), but a system of the size and complexity of Genie requires a supporting semantic framework. 2

4. 3 What is GENIE?  Grid ENabled Integrated Earth system model.  Investigate long term changes to the Earth’s climate (i.e. global warming) by integrating numerical models of the Earth system.  e-Science aims:  Flexibly couple together state-of-the-art components to form unified Earth System Model (ESM).  Execute resultant ESM on a Grid infrastructure.  Share resultant data produced by simulation runs.  Provide high-level open access to the system, creating and supporting virtual organisation of Earth System modellers.

5. 4 GENIE model framework 3D atmosphere 3D ocean 2D sea ice Atmospheric CO 2 2D land surface Land biogeochemistry Ocean biogeochemisty Ocean sediments 3D ice sheets

6. Previous Work  Separation of the code into pieces - Layered representation of the application.  Wrapping GENIE as ICENI binary component.  Delivery of Web-based System (portal) to allow a virtual organisation of environmental scientists to create and manage simulations at a high-level.  Submission of multiple sweep experiments to the portal, exploiting multiple resources over a Condor pool.  Implementation of database management system to allow scientists to share, access and visualise data produced by simulation runs. 5

7. Currently in GENIE  genie.F is a glorified metafile containing if-then-else statements.  Fortran routines act as “wrappers”, IN/OUT annotation.  Life cycle phases scattered, i.e. igcm_adiab.F and initialise_atmos.F.  Shared memory, file based I/O.  Complex semantics (e.g. grid interpolations).  Hard-coded time steps. 6

8. 7 OLOGEN Ontology for GENIE:  Classes  Relations

9. Entities (1) OLOGEN Class Hierarchy:  Abstractions - What is the abstraction granularity?  Component  Module (atmosphere, land, ocean etc)  Model (IGCM, c-Goldstein, etc)  Wrapper ∶  Routine  Function (…)  Interface  Data Flow Type (IN, OUT, INOUT, Control)  Model Scope (Common, Model-Specific)  Binding (Late, Current)  Layer (Local, Wrapper) ∶ 8 is part of

10. Entities (2) 8

11. 10 Relations

12. 11 Higher-order Functions Co-ordination forms  The Wrap Skeleton  The Thread Coupler Skeleton  Distribution Model  Same-host  Distributed  Parallel

13. 12 The Wrap Skeleton Wrap( c,list of wrapper input interfaces, list of wrapper output interfaces, list of local input interfaces, list of local output interfaces, list of control interfaces)=C

14. 13 The Thread Coupler Skeleton [SCL, Darlington93] IterFor koverall { IterFor n step A IterFor m step B where step A = threadA.execute threadB.notify(interfaceB) step B = threadB.execute threadA.notify(interfaceA) }

15. Nested Coupling igcm_land_surflux pass results 3 pass lowest level values to ocean & land igcm3_adiab flag_land not igcm_ocean_surflux igcm_ocean_blaye r flag_goldsteinocean and flag_goldsteinseaice not check_fluxe s_surf_l flag_checkfluxes_surf not igcm_land _blayer genie_land_surflux igcm3_diab pass results 2 land_niter_tim katmos_int ocean_niter_tim 14

16. 15 Conclusions  Use of reverse engineering approach to research on component coupling frameworks:  Analysis of a monolithic piece of code such as GENIE.  Extraction of interface and model coupling semantics.  Development of OLOGEN, an ontology for capturing such semantics and validating the development.  Co-ordination via functional skeletons.

17. 16 Acknowledgments  We would like to thank Andrew Price, Dan Lunt, Gethin Williams, Ian Rutt, Andrew Yool, Neil Edwards, Tim Lenton and John Shepherd for their assistance and guidance.  Acknowledgments also go to the GENIE investigators: Prof. Paul Valdes (Reading), Prof. John Shepherd (SOC, Southampton), Prof. Andrew Watson (UEA), Prof. Melvyn Cannell (CEH Edinburgh), Dr. Anthony Payne (Bristol), Prof. Richard Harding (CEH Wallingford), Prof. Simon Cox (SReSC) and Prof. John Darlington (LeSC).  Reference: [SCL, Darlington93] J. Darlington, Y. Guo, H. W. To, and J. Yang. Functional skeletons for parallel coordination. In Seif Haridi, Khayri Ali, and Peter Magnussin, editors, Euro-Par'95 Parallel Processing, pages 55-69. Springer-Verlag, August 1995.