A Multi-Agent System on Replication of Redundant Systems

• Main Authors: Yu-ChiLin, 林侑琦
• Other Authors: Tzone-I Wang
• Format: Others
• Language: zh-TW
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/16028207113326773521

碩士 === 國立成功大學 === 工程科學系專班 === 98 === With rapid advances of computer and information technology, an increasing number of enterprises are using very large computer information systems to store enterprises managing data. Once a system-crash occurred owing to natural disasters or human errors, such as typhoons, earthquakes, or fires, quick recovery from disasters that brings the systems to live will minimize the loss for enterprises. Accordingly, the backup of data as well as systems becomes an important issue. However, since online systems cannot be shut down for maintenance frequently, the number of application software and their configurations on the redundant server may become asynchronous with the online systems after a while. Furthermore, current backup mechanism is mostly developed for data replication, which cannot be used to replicate application software. So when a disaster occurs, switching to the replicated servers will not bring the whole information systems right back to the latest status before the crash. This paper establishes a mechanism based on software agents for replicating application software of online servers to replicated servers, using characteristics of autonomy and social communication of software agents. The software agents on the online servers monitors the changes of application software and records all the steps of installations and configurations, and then deliver recorded scripts to the software agents on the replicated servers, which then initiate the automatic installations and configurations with the scripts. So when an online server completes the installation and configuration of an application, its replicated server will automatically finish the same installation and configuration of the same application. The synchronizations of applications on both online and replicated servers can then be accomplished.