Research And Design Of Server Cluster System For Mobile Social Network

With the rapid development of wireless communication technology and intelligent terminal equipment, a new form of social networks,mobile social networks have been rising. According to the discovery, at present the research on mobile social networks can be divided into two directions, one is the extension and expansion of online social network services in mobile intelligent terminals, the other is the analysis and design of message routing strategy in mobile social networks. Between them, the former is mainly focused on the service application, and the latter is mainly devoted to the development of the forwarding and delivery rules of user message in different scenarios.The research content of this paper belongs to the first branch of the research of mobile social networks. In-depth study of the specific application requirements of users in mobile social networks, this paper designed and developed a mobile social network server based on openfire, which is a open source instant messaging server. The server function is composed of two parts. The basic function is mainly to meet the users’ basic social needs, the characteristic functions provide two services with the feature of mobile social networks, one is a location service based on the relationship, the other is a recommended method based on similarity of three dimensions. In order to validate the friend recommendation method, used the user data obtained from the Four Square, a mobile social networking platform, to test.In order to provide better stability and efficient mobile social network services to more users, this paper introduced the LVS cluster technology, which made the single service node extend to the cluster structure, form the server cluster system oriented mobile social networks. According to specific application scenarios, two shortcomings of the cluster system were analyzed, one is "IO" communication bottleneck problem between node servers and central database due to the shared database; the other is low efficiency of request processing due to the LVS default load balancing algorithm-- the WLC algorithm can’t dynamically adjust the weights and accurate assessment of node servers.In order to solve the two problems mentioned above, this paper proposed two alternatives to optimize the performance of server cluster system. The first is a data storage transformation based on the redis-cluster. The second is a load balance algorithm for mobile social networks based on WLC algorithm, which is the default load scheduling algorithm in LVS. Finally, this paper used a small cluster built by myself to test the performance of the server cluster system. From the experimental results, we can see that the optimization scheme presented in this paper, can effectively enhance the performance of server cluster system oriented mobile social networks. Between them, the load balanced optimization scheme had a better effect than the data storage optimization scheme.