The Parallel Common Component Architecture in SCIRun2

by
Keming Zhang

Advised by
Steve Parker

In recent years, component technology has emerged as a successful approach for large-scale commercial software development. Existing commercial component models, such as Java Beans, Microsoft COM and OMG CORBA are not suitable for large-scale scientific computation, because they lack many high-performance computing features. The Common Component Architecture (CCA) Forum has defined a standard component architecture for high performance parallel computing. CCA defines a minimal set of standard interfaces that a high-performance component framework should provide to implement high-performance components. This standard promotes interoperability between components developed by different teams across different institutions. However, parallel CCA is still an open research area. This project is trying to define the semantics for parallel components support in the SCIRun2 environment. Moreover, it creates and evaluates the design of parallel CCA components in SCIRun2. Our parallel CCA is based on a Parallel Remote Method Invocation (PRMI) mechanism. This software architecture will encompass both distributed-memory systems and shared memory systems. It will allow different parallel machines to be combined within the same application, and in some cases will ease the application portability between the different memory systems. In this way, the parallel CCA architecture will make the component technology more suitable for large-scale scientific computing.