Research On Detecting Technique For Cluster Monitoring System

As network application becomes more popular, network servers are required with higher performance. Compared with traditional high performance computer, cluster system has become prevailing network server because of its excellent cost-performance ratio and high scalability. A cluster server is composed of many computing nodes, so it is very important to monitor every node for the safe and efficient running of the cluster server. The goal of this paper is to do deep research on the cluster detecting technique and interference-reducing technique:Under the centralized monitoring framework, a cluster detecting method is proposed. By detecting file security, network port's state, user action and account security, the security detecting function is promoted in the cluster monitoring system and the security detecting problem is solved in the cluster system. This paper emphatically presented a rule-based file detecting technique and its implementation, and the function is boosted up in the cluster monitoring prototype system.Subsequently, aiming at the new interference brought by the detecting function and using the current methods for reference, this paper provides a multi-layer data filtering model and a content-based data filtering algorithm to reduce the interference by cutting down the transfer quantity of data. Additionally, this paper presents a connection-based network control algorithm to reduce the interference by cutting down the useless TCP connection. In order to validate the model and algorithm and test their efficiency in the application, experimental schemes are implemented on the cluster monitoring prototype system. The results show that the expected performance is achieved and the filtering algorithm and network control algorithm can reduce the new interference brought by detecting function as well as the traditional interference resulting from performance monitoring.The detecting technique presented in this paper has been applied in the cluster monitoring prototype system of NNS(New Network Server) with excellent performance in monitoring the system's security. The research provides good application value for detecting cluster security and reducing the interference.