Specification of a Scilab meta-computing extension

Scilab is an application for numerical algebraic computations developed in the Meta-2 INRIA project. This program contains a toolbox allowing the realization of complex simulations with high transparency for the user. Nevertheless, computation times and data sizes may be quite large. Thus it seems interesting to provide an access to the power of parallel computers to the Scilab users. A parallel version is already available, Scilab/sub /spl par//, but it requires a direct connection to a parallel computer or a network of workstations. Moreover the parallelism is explicit and the users must have a good skill in parallelism. This is why we propose a meta-computing extension of Scilab in order to keep a high level of transparency according to the user. Thus, a non-specialist user will be able to make parallel computations without knowing how the data are distributed and what parallel server has been chosen to make the computations. Moreover this must be possible from a simple Scilab session running on a classical workstation. The client/server model used in this platform is a classical example of the use of an object-oriented distributed system, like those following the CORBA norm. We describe the rules for the use of such a tool and also the set of CORBA services which have to be designed to realize this task. In particular we point out the need for a localization service and a trader. Some problems like data persistence over the servers and asynchronism between independent Scilab instructions must be studied.