The Self-organization Detection Technology Based On Gossip Protocols

The computer network and distributed system has been applied more and more widely, and the scale and complexity of the network has increased exponentially. While the high reliability and performance requirement appears more and more important, to achieve and maintain it is difficult, especially in some key fields (such as banking, medical treatment, military and aerospace, etc.), People put forward higher availability and reliability requirements for important network nodes (such as the server group). So nodes fault monitoring technology arises at the historic moment, and builds the solid foundation for effective fault-tolerant.As the large-scale and dynamic characteristics of network deepen further, the adaptive ability and expansibility of the nodes monitoring becomes very important. The traditional fault detection and news spread mode will cause delay, network congestion and so on, which will decrease monitoring performance, even becomes extra burden of effective service. Based on the present research of the fault detection and referencing the features of the self-organizing network, this paper comes up with a self-organization neighborhood detection method based on gossip protocols. This topic relies on the distributed software reliable monitoring research projects of Chongqing Natural Science Fund, and it contains two aspects about node monitoring section, one is Self-organizing field partition and news spread mechanism, which is the research focus of this paper, the other is data collection and failure analysis of key nodes. The following is the concrete research contents.①It summarizes the characteristics of self-organization network and research methods, discusses Gossip protocols, the P2P communication and techniques of software development.②It explores NAT type testing and penetration technology of P2P communication, and designs a UDP penetration method between the private networks in allusion to common NAT type, to ensure the effective peer-to-peer communication in WAN.③It researches the design ideas and implementation methods of self-organization neighborhood mechanism, references a lot of specific applications and design thoughts of self-organization networks, constructs the detection neighborhood groups which use agent node as center, and designs the multi-forwarding message broadcasting method based on Gossip protocols. ④It researches the manage mechanism of self-organization neighborhood networks. With regard to the situations of the nodes failure and abnormal, new node's joints and log out etc, it expands a three-layer management model based on the neighborhood divided system, which can ensure the normal operation of the self-organization neighborhood mechanism. At last, it discusses the occasions of neighborhood's repartition.This paper develops and achieves the above content in the Linux system. The test data and result analysis show us: this mechanism can not only reduce the cost of networks fault detection significantly, but also reduce the chance of the backbone network's delay and congestion effectively, hence enhances the detection performance.