Logo image
Runtime Compositions and Optimizations of Parallel JavaBean Programs on Clustering Environments
Conference paper

Runtime Compositions and Optimizations of Parallel JavaBean Programs on Clustering Environments

Cheng-Wei Chen, Chung-Kai Chen and Jenq Kuen Lee
Proceedings of the International Conference on Parallel and Distributed Processing Techniques and Applications, Vol.4, pp.1919-957
2000

Abstract

distributed object management;Java;object-oriented programming;optimising compilers;parallel programming
We describe ongoing research work to provide a software infrastructure for the runtime composition and optimization of parallel JavaBean programs. The system includes several essential components. First, a software module called component annotator, is employed. An application-level JavaBean program (component) can be annotated to produce trace information at runtime. This information can be used for further program optimizations. Second, application components are integrated with a container component proposed by us before execution. In the runtime execution, the container component can act as an active program unit to analyze the trace information, perform component optimization, produce the new version of a chosen application component, and perform runtime composition of application components by replacing the original application component with the new component. Our system works on the component level, deals with parallel optimization issues, and optimizes the programs by utilizing application characteristics. In an application scenario, our system can be used to replace a component from parallel JavaBean programs by a new version of a component at runtime for efficiency. We illustrate such an optimization for a fluid application component with unstructured grids and sparse computations on an IBM SP-2. The system provides an important framework and direction in the optimization of parallel Java components at runtime

Metrics

1 Record Views

Details

Logo image