Parallelisation of an Ultra-short Range Meteorological Prediction System by P-GRADE development environment

Lovas Róbert <rlovas@sztaki.hu>

MTA SZTAKI

Horváth Ákos dr. <horvath@met.hu>

Országos Meteorológia Szolgálat

Mesoscale Analysis and Ultra-short Range Prediction System (MEANDER) developed by the Hungarian Meteorological Service has a crucial task in the protection of life and property; based on computations using all the available meteorological observations the system can issue weather warnings which are essential for storm warning at Lake Balaton, for aviation, etc.

MEANDER consists of some computational intensive procedures that cannot be executed on time without an efficient parallelisation. MTA SZTAKI and University of Miskolc developed the P-GRADE graphical programming environment that has been used for the parallelisation of the sequential FORTRAN and C++ code, supporting the entire life cycle of parallel program development.

In this paper we introduce briefly some essential ultra-short range prediction procedures that have been parallelised by P-GRADE program development environment as well as the way of parallelisation during the design, debugging and performance analysis phases. Furthermore, the results of executions on supercomputer and cluster environments are also presented in order to illustrate the efficient usage of P-GRADE environment on both platforms.