Research On Kernel Level Cluster Communication Monitoring System

With the development of the cluster system, the scale of cluster communication network is becoming larger, and the complexity of it is increasing. So the research on the cluster communication monitoring system is very important to guarantee the system robustness,debug the system software and improve the performance. This dissertation analyzes the related problems of cluster communication monitoring system, including the basic theories and implementation. We propose a manager/agent model and a kernel level monitoring strategy, and describe a kernel level monitoring system based on the BCL-4 basic communication protocol.After introducing the basic concepts and architecture of cluster and cluster communication system, this dissertation analyzes the architecture and monitoring level of cluster communication monitoring system. The manager/agent model is given, and the concepts of user level monitoring, kernel level monitoring, card level monitoring are presented. The character, merit and difficulty of kernel level monitoring are also discussed. Then the dissertation concentrates on the design and implementation of a kernel level cluster communication monitoring system based on the BCL-4 basic communication protocol.After introducing the Dawning cluster communication system and the BCL-4 communication protocol, the structure and the design of the monitoring system, the kernel level monitoring protocol, and the key techniques of system implementation are described in detail.Finally, the dissertation analyzes the monitoring system, concludes the work that we have finished, and gives some aspects of the research work that we might do next.