The development of large parallel computing systems - systems with hundreds or thousands of processors - is increasingly focused on designs involving clusters of multiprocessor shared memory (SMP) nodes connected by a network. Nodes typically consist of several (from 2 to 16+) commodity microprocessors with shared access to an amount of memory local to the node.

We propose that the clustered SMP design offers unique characteristics that may be exploited to support applications composed of two or more distinct jobs with complimentary resource access patterns. One example of this type of application is computational steering.

We propose to study the design and performance characteristics inherent in computational steering applications on clustered SMP machines. As a first step, work to be undertaken as part of this proposal will involve coupling an existing simulation code with an existing scientific visualization code. These codes will be incrementally extended to share data and control parameters. This sequence of steps will enable comparative analysis of the overall efficiency of the tandem codes in environments similar to traditional parallel systems, and in clustered SMP systems.