Research And Implementation Of Key Technologies Of Distributed Communication Cluster Based On XMPP Protocol

With the rapid development of computer netwo k and communication technology, more and more people get information through the Internet, a large number of users enter the Internet world.Increased and more complex demand for the various information centers and data processing center occur. The server must accept reasonable client requests and responds in a short period of time, besides the server must remain highly available and reliable. For example, Tencent Wechat,QQ and other network communication services, will withstand a huge traffic every day. Such a huge data traffic and complex transmission requirements, there is an urgent need for an information communication server of a rapid response capability, high scalability, high availability. In the complex mobile platform environment, how to effectively and quickly respond to client requests and make full use of the resource capacity of the back-end server nodes has become a difficult and important issue which affects the distributed communication cluster architecture.With the real distributed communication cluster environment based on XMPP protocol, this paper researches the key technology of instant messaging, distributed communication cluster, load balancing and secure communication. What’s more, this paper deal with the scheduling allocation problem in the communication environment based on XMPP protocol. Load balancing algorithm is the core problem of load balancing server which is the front-end of the XMPP communication server cluster.It is the prerequisite to realize reasonable dispatching. This paper analyzes the problems of existing load balancing algorithm. To address the problems in the mobile platform based on XMPP protocol, this paper presents a prediction method of the server load based on Classification and Regression Tree and KNN machine learning algorithm. The algorithm makes up time series of load by gathering every node’s load information comprehensively and analyzes the time series, and then divides the server nodes into three regions, uses different scheduling strategy in different regions. Simulation experiments and tests show that compared with Nginx round robin and fair algorithm, this proposed prediction algorithm decrease by 25% in connection response time, and increase by 1.3 times in connection establishment, dynamic scheduling strategy makes the communication server cluster has a greater network throughput and more robust adaptability for mobile platform.At the same time, this paper designs and realizes the distributed communication cluster system based on XMPP protocol. It introduces the requirement analysis and design principle of the system in detail and divides the system into modules, designs and implements each module.The key technologies include server cluster, secure communication,database and load balancing. The introduction of secure communication technology takes the internal and external security of the cluster into account and breaks and improves several key points in the load balancing algorithm to improve the usability, scalability and fast response capability of the cluster system.