Multi-agent Cluster System for Optimal Performance in Heterogeneous Computer Environments

In an environment where various personal computers exist together, computers may have different operating systems (OS) and various specifications. In this case, an OS which is not suitable for constructing a PC cluster may be included in that environment and the performance of PC cluster is influenced by personal computers with low specification. In this research, automatic OS decision function, grouping function and automatic start scheduling function are suggested to reduce the influence given by these problems. These functions are realized by multiple agents. In addition, by showing the difference of performances between diskless nodes and non-diskless nodes, we verified the effectiveness of automatic OS decision function.