A Java prototype implementation of coordination for heterogeneous, distributed, and parallel programming.

摘要

Heterogeneous distributed parallel programming skills are in rapidly rising demand for developing applications in emerging fields such as MRI spin relaxometry, gene sequence analysis, climate modeling, and molecular modeling of potential bioactive compounds that require massive amounts of computation. However, heterogeneous distributed and parallel programming is not a trivial task for programmers. Programmers are overwhelmed to program at a low level with specific target machines or are faced with unfamiliar programming models. In this thesis, we introduce a framework to support heterogeneous distributed programming in a sequential programming language Java for parallelism. The framework has four layers (application layer, coordination layer, middleware layer, and platform layer). We implemented a Java prototype of the framework using the Interacting Processes (IP) algorithm and CORBA as a communication infrastructure. The IP algorithm is used for the support of parallelism via coordination among processes and CORBA is used as a communication middleware to support data exchanges across different platforms. An example of the dining philosophers problem is illustrated to demonstrate heterogeneous distributed parallel programming using the framework.