A Multi-Agent System based on Specific-Profile for Efficient Distributed Computing

We propose the multi agent system based on profiling specifics for the distributed system, SPMaS. The job scheduling and the path finding on the distributed computing are important factors for the reliable system and running time reduction. The factors which affect the entire performance of scheduling and path finding are node's capability(CPU, memory), network state(speed, traffic frequency) and usage patterns(average available resource, connection lasting time, job request frequency, job execution frequency and error frequency). SPMaS analyses the characters of node by using agent and composes node history. The node history helps system allocate the best-suitable job for each node during job scheduling. In addition, managing groups based on the node history makes effective handling for the load balance and defect of transmission. Key-Words: Multi Agent System, Mobile Agent, Specific Profile, Distributed Computing