Agent-Based Middleware for Grid Computing

WADE system is a dynamism-adapted campus computational grid using agents to mask resource heterogeneity and distribution. In this work, an agent-based method is given for parallel computing on the campus grid. The framework and primarymodules of the agent-based middleware of WADE system are described in detail. Then, the steps of compiling and execution on heterogeneous nodes using agents are presented. Finally, agents are implemented in Java. Here, agents act as the middleware of grid computing, which coordinates several parallel programming environments like MPI and PVM to support heterogeneous parallel computing, and provides uniform service interfaces for clients. Thus, the usability of the grid system is improved distinctly.