Parallel Remote Visualization for the Grid

The availability of a remote visualization service is widely recognized as a key aspect for Grid based scientific virtual organizations. The design of this kind of service can be considered from two different but complementary points of view: the deployment of the service as Grid middleware, and the design of effective and efficient building blocks for the visualization service. In this work we are concerned with the design of efficient and effective building blocks for remote visualization. We will take in consideration both algorithmic and programming aspects. The algorithms design analysis shows the fundamental role of parallel processing, and the need of structured parallel programming in order to manage the design and implementation of complex parallel components. At the same time the use of a high performance component based model is necessary to obtain the desired design flexibility, integration and interoperability. We consider these aspects reporting our experience in the design and implementation of a visualization pipeline. In particular the design of a component for isosurface extraction is considered in details and its implementations using a high level structured parallel programming environment, and a Common Component Architecture (CCA) compliant framework are discussed. Experimental results are presented and performances obtained with a C-MPI based implementation are compared with performances of the structured parallel programming and component-based versions.