Supercomputer centers offer high capacity computer resources destined to solve complicated computer problems (the so-called  "great computational challenges").

In recent years, the efficiency of the processors has grown steadily in accordance with Moore's law, and new massive parallel computer achitectures have been developed. However, no solution has been found to many other problems in different areas, since they require extremely long computational times. Some others could not even be dealt with at all due to the lack of computational resources such as memory capacity or permanent data storage.

Like in many other areas, the only way to solve these problems is to cooperate and share the resources among the different institutions. Grid computing provides the  mechanisms needed to achive this association of computational powers. One of the most famous grid projects is Globus, where a group of software components, which would later become de facto standards, were developed.

Within the different Grid projects, many different testbeds appeared, whose function was to experiment and to make the most of this technology. However, most projects related to grid (such as Datagrid, the european project) focus on sequential tasks, which was called "high processing capacity", and a few others focus on parallel tasks within grid. This is what has happened in the simulation outline "Cactus" and the european project "Crossgrid". These parallel applications can benefit from the latest progress in wide area networks as regards bandwidth and reliability, as well as from the recent introduction of service quality policies for these infrastructures.

Two supercomputer centers in Spain ("Centro de Supercomputación de Galicia"-CESGA- and "Centre de Supercomputació de Catalunya"-CESCA-) took the first step in that direction, leading a series of experiences which will help to know whether these technologies are practically useful for parallel applications, and to determine the problems and benefits that the use of Globus toolkit can present.      

Aims

The principal aims of this experience are:      

• To create a grid computing environment between the two independent supercomputer centers.
• To demonstrate that the practical sequential problems can be considered within this grid environment, and that load balancing can be achieved with no changes in codes or reconfiguration of systems.
• To demonstrate that the parallel codes can be executed within this grid environment with no changes in codes or reconfiguration of systems.
• To determine how the network heterogeneity and the communication team can have an influence on this kind of problems, and to determine the reliability, availability and capacity to render service (RAS) of the new generation of research networks in wide areas.

Behind all these important aims, there is a common effort to provide exclusive and highly efficient computer resources that help to solve scientific problems that require wider computer capacities.